Courseiva

CCNA Monitor Optimize Analytics Solution Questions

75 of 102 questions · Page 1/2 · Monitor Optimize Analytics Solution topic · Answers revealed

1
MCQmedium

You are configuring a Spark pool in Microsoft Fabric for a workload with highly variable data volumes. You want to ensure that the pool can handle peak loads without over-provisioning resources during quiet periods. Which setting should you adjust?

A.Dynamic Allocation of executors
B.Autoscale on the Spark pool
C.The Max Memory per node setting
D.The default Spark version
AnswerB

Autoscale allows you to define a minimum and maximum number of nodes for the Spark pool. The Fabric orchestrator automatically monitors the workload and scales the pool size within these boundaries. This ensures that resources are available for heavy processing while minimizing idle capacity and associated costs when the pool is quiet.

Why this answer

Autoscale on the Spark pool dynamically adjusts the number of nodes based on workload demand, allowing the pool to scale up for peak data volumes and scale down during quiet periods. This directly addresses the requirement to handle peaks without over-provisioning.

Exam trap

DP-700 often tests the confusion between Spark's dynamic allocation (executor-level, within fixed nodes) and Fabric's autoscale (node-level, adjusts pool size), leading candidates to pick the wrong scaling mechanism.

How to eliminate wrong answers

Option A is wrong because dynamic allocation of executors operates within the nodes already allocated to the pool; it does not add or remove nodes to match cluster-level demand, so it cannot handle peak loads that exceed the pool's node capacity. Option C is wrong because setting max memory per node is a static configuration that does not adapt to variable workloads and can cause over- or under-provisioning. Option D is wrong because the default Spark version is unrelated to resource scaling and does not affect capacity management.

2
Multi-Selectmedium

Which THREE features are provided by the Fabric Capacity Metrics app for monitoring performance?

Select 3 answers
A.Identifying which users are consuming the most capacity.
B.Modifying the underlying SQL Server instance settings.
C.Monitoring interactive versus background operation usage.
D.Analyzing the consumption of specific Fabric items.
E.Automatically scaling the capacity SKU.
AnswersA, C, D

The app tracks resource consumption by user, which helps administrators identify if specific users or departments are running inefficient reports or processes that lead to capacity throttling or high utilization, facilitating better resource management and potential user training on efficient query habits.

Why this answer

The Fabric Capacity Metrics app is a monitoring tool that surfaces capacity consumption data across the tenant, so option A is correct because it lets administrators identify which users are consuming the most capacity, helping pinpoint heavy consumers and potential bottlenecks. Option C is correct because the app distinguishes between interactive operations (such as user-driven queries and reports) and background operations (such as scheduled refreshes and data pipelines), which is essential for understanding how capacity is being used. Option D is correct because the app provides item-level breakdowns, allowing analysis of consumption for specific Fabric items like semantic models, reports, and dataflows.

Option B is incorrect because the app is read-only for monitoring and does not modify SQL Server instance settings. Option E is incorrect because automatic scaling of the capacity SKU is not a feature of the app; scaling is handled separately through capacity settings or automation tools.

Exam trap

DP-700 often tests whether candidates over-attribute capabilities to the Capacity Metrics app, tempting them to select options about modifying settings or auto-scaling — features the app does not provide.

3
Multi-Selectmedium

You need to monitor the health and performance of your Microsoft Fabric environment. Which two components should you primarily monitor to identify bottlenecks in data movement?

Select 2 answers
A.On-premises Data Gateway performance
B.Pipeline activity and duration logs
C.OneLake file storage size
D.User interaction counts on Power BI reports
E.Workspace user list
AnswersA, B

The gateway is the conduit for moving data from on-premises sources to Fabric. Monitoring its CPU, memory, and throughput is essential to ensure that data transfer is not being delayed by local network constraints or insufficient gateway capacity, which can often be the hidden cause of pipeline slowdowns.

Why this answer

Option A (On-premises Data Gateway performance) is correct because the gateway is the bridge that moves data between on-premises sources and Fabric, so its CPU, memory, and concurrent-connection metrics directly reveal throughput bottlenecks during data movement. Option B (Pipeline activity and duration logs) is correct because pipeline run history, activity durations, and copy throughput metrics expose where data movement stages slow down or fail. Option C is not a primary bottleneck indicator since OneLake storage size reflects capacity consumption, not movement speed.

Option D measures report consumption behavior, not data movement performance. Option E is an access-control artifact and has no bearing on data movement throughput.

Exam trap

DP-700 often tests the distinction between capacity/storage metrics and performance metrics, tricking candidates into selecting storage size or user counts as indicators of data movement bottlenecks.

4
MCQhard

You are tuning a Microsoft Fabric Spark notebook that reads a large Delta table in a Lakehouse and writes aggregated results back to the same Lakehouse. Job history shows the write stage spends significant time rewriting files, and you notice the target table has accumulated many small files over months of incremental appends. You need to reduce write amplification for future runs. Which approach should you implement?

A.Enable Delta Lake change data feed on the target table to capture incremental changes during writes.
B.Convert the target table to a Parquet-only folder and drop the Delta transaction log to eliminate commit overhead.
C.Increase the number of shuffle partitions in the Spark session so that each task writes a smaller slice of data.
D.Schedule periodic OPTIMIZE with ZORDER on the target Delta table and enable optimized writes in the Spark session configuration.
AnswerD

Combining OPTIMIZE with ZORDER compacts the accumulated small files and co-locates related data, while optimized writes reduce the number of files each run produces. Together they attack both the historical fragmentation and the ongoing write amplification, which is precisely the described problem of many small files degrading write performance.

Why this answer

Write amplification in Delta comes from repeatedly creating many small files, so the durable fix combines compaction of existing files with preventing future fragmentation. OPTIMIZE with ZORDER compacts and clusters data, and optimized writes reduce file count per run. Dropping the Delta log, inflating shuffle partitions, or enabling change data feed either removes necessary guarantees or worsens fragmentation, none of which reduces the write cost.

Exam trap

The trap here is assuming that more shuffle partitions always speeds up writes, when in Delta each partition typically produces its own small output file.

5
MCQeasy

A data engineer is using the Microsoft Fabric Monitoring hub to track the status of data pipelines. They need to quickly identify which pipeline runs have failed in the last 24 hours. What should they do?

A.Open each pipeline individually and check the run history for failures.
B.Use the Fabric Capacity Metrics app to view failed pipeline runs.
C.Navigate to the Monitoring hub, filter by 'Status' = 'Failed', and set the time range to the last 24 hours.
D.Query the Fabric REST API for pipeline run history and filter by status.
AnswerC

The Monitoring hub in Microsoft Fabric provides a centralized view of all activities, including pipeline runs. By applying a filter for failed status and setting the time range to the last 24 hours, you can quickly isolate the failed runs. This is the most direct and efficient method to identify recent failures and take corrective action.

Why this answer

The Monitoring hub is the primary tool in Microsoft Fabric for monitoring activity across items, including pipelines. It allows filtering by status and time range, making it straightforward to identify failed runs within a specific period. Other tools like the Capacity Metrics app or REST API serve different purposes and are less efficient for this quick, visual inspection task.

Exam trap

The trap here is confusing the Monitoring hub with the Capacity Metrics app; the latter is for capacity-level performance, not for tracking individual pipeline run statuses.

6
MCQmedium

You are using the Microsoft Fabric Capacity Metrics app and notice that your capacity is frequently hitting the limit for 'Interactive' operations. What should you do first?

A.Immediately scale up the capacity SKU.
B.Use the metrics app to identify top-consuming items.
C.Clear the browser cache for all users.
D.Delete all unused reports in the workspace.
AnswerB

The first step in any capacity optimization process is to identify which items are causing the high resource usage. The Capacity Metrics app provides drill-down capabilities to pinpoint the specific datasets or reports, enabling a targeted approach to optimization rather than a broad, costly infrastructure change.

Why this answer

The correct first step is to use the Microsoft Fabric Capacity Metrics app to identify the top-consuming items, because it provides per-item and per-operation breakdowns of Interactive vs Background consumption. Scaling up the SKU without knowing which artifact is causing the overload can waste cost and may not resolve the root cause. The metrics app surfaces the specific datasets, reports, or notebooks driving Interactive usage so you can optimize or throttle them before resizing.

Exam trap

DP-700 often tests the reflex to scale up capacity immediately, but the exam expects candidates to first use the Capacity Metrics app to identify the top-consuming items before resizing.

How to eliminate wrong answers

Option A is wrong because immediately scaling up the capacity SKU is a costly reactive action that does not diagnose which item is consuming Interactive capacity, and the problem may persist if a single inefficient report is the cause. Option C is wrong because clearing browser caches has no effect on Fabric capacity consumption metrics, which are server-side. Option D is wrong because deleting unused reports is a guess; the metrics app must first identify whether unused reports are actually contributing to the Interactive limit before deletion is justified.

7
Multi-Selecteasy

You need to quickly identify all failed Data Factory pipeline runs and Spark jobs across multiple workspaces in your Microsoft Fabric environment. Which TWO actions should you perform in the Fabric Monitoring Hub?

Select 2 answers
A.Apply a filter for the 'Failed' status.
B.Use the 'Item type' filter to select Pipeline and Spark job.
C.Export the logs to an Azure Log Analytics workspace.
D.Schedule a daily refresh for the Monitoring Hub dashboard.
E.Configure an alert in the Capacity Metrics app.
AnswersA, B

Filtering by status is the most efficient way to isolate jobs that require immediate attention. In a busy environment, this reduces noise and allows engineers to focus on troubleshooting errors. The Monitoring Hub allows this filter to be applied globally across different types of Fabric items simultaneously.

Why this answer

The Monitoring Hub is designed for quick, interactive triage of activity across workspaces, so applying a filter for the 'Failed' status (option A) immediately narrows the view to only unsuccessful runs, which is exactly what the scenario requires. Using the 'Item type' filter to select Pipeline and Spark job (option B) restricts the results to the two artifact types named in the requirement, so failed Data Factory pipeline runs and Spark jobs are isolated across all workspaces. Together these two filters let you rapidly identify the failed items without exporting or scheduling anything.

Exporting logs to an Azure Log Analytics workspace (option C) is a diagnostic/retention action, not a quick identification step in the Monitoring Hub. Scheduling a daily refresh for the Monitoring Hub dashboard (option D) is not a Monitoring Hub capability and would not filter failures. Configuring an alert in the Capacity Metrics app (option E) monitors capacity consumption, not individual failed pipeline or Spark job runs.

Exam trap

The trap is confusing the Monitoring Hub with the Capacity Metrics app; the former is for run-level monitoring, while the latter is for capacity utilization and throttling.

8
MCQhard

You are monitoring a Microsoft Fabric Data Pipeline that ingests data from an on-premises SQL Server into a Lakehouse. The pipeline occasionally fails during the copy activity with a 'timeout' error. You need to capture detailed diagnostic information, including the exact error message and the activity run ID, to troubleshoot the issue. Which monitoring feature should you use?

A.Azure Monitor integration with Fabric
B.The pipeline run history in the Fabric workspace monitor
C.The 'Monitor' hub in Microsoft Fabric, drilling into the specific activity run
D.The Data Pipeline's built-in 'Logging' settings with diagnostic logs enabled
AnswerC

The Monitor hub provides a centralized view of all pipeline runs and allows you to drill down into individual activity runs. For a failed activity, it displays the error message, activity run ID, and other diagnostic details, which are essential for troubleshooting timeout errors and identifying the root cause.

Why this answer

The Monitor hub in Microsoft Fabric provides detailed activity run information, including error messages and activity run IDs, which are critical for diagnosing intermittent failures like timeouts. By drilling into the specific activity run, you can see the exact error and correlate it with other events. Other options either lack the required detail or require additional configuration that may not be readily available during troubleshooting.

Exam trap

The trap here is confusing high-level pipeline run history with the detailed activity-level diagnostics available in the Monitor hub.

9
MCQeasy

You want to monitor the health and performance of your Microsoft Fabric pipelines. Which tool should you use to view detailed execution logs and identify bottlenecks in your data movement?

A.Power BI Desktop
B.Fabric Monitoring Hub
C.Azure Resource Graph
D.Microsoft Purview
AnswerB

The Fabric Monitoring Hub is the primary interface for tracking the health and performance of all Fabric operations, including pipelines. It provides detailed logs, error messages, and execution metrics that allow you to pinpoint specific activities causing delays in your data integration workflows.

Why this answer

The Fabric Monitoring Hub provides a centralized view of all activities, including pipeline runs, with detailed execution logs, durations, and error messages. It allows you to drill into individual pipeline runs to identify bottlenecks and performance issues in data movement.

Exam trap

DP-700 often tests the misconception that Power BI Desktop or Azure tools can monitor Fabric pipelines, when the correct tool is the Fabric Monitoring Hub.

How to eliminate wrong answers

Option A is wrong because Power BI Desktop is for authoring reports, not for monitoring Fabric pipelines. Option C is wrong because Azure Resource Graph is for querying Azure resources, not Fabric pipelines. Option D is wrong because Microsoft Purview is for data governance and compliance, not for pipeline performance monitoring.

10
MCQmedium

Your team runs a Fabric pipeline that copies data from Azure SQL Database into a Lakehouse every two hours. A data engineer reports that one recent run took far longer than usual and wants to know which activity in the run consumed the most time. You need to identify the slowest activity and inspect its duration and status. What should you do?

A.Open the pipeline run in the Monitoring hub, then drill into the run's activity-level details to compare durations.
B.Review the Capacity Metrics app and sort items by the timepoint with the highest consumption.
C.Enable verbose logging on the Azure SQL Database source and read the server-side trace output.
D.Query the Lakehouse SQL analytics endpoint for a log table that records each pipeline activity.
AnswerA

The Monitoring hub exposes pipeline runs and lets you drill into individual activity executions within a run. Each activity shows its own status, start time, and duration, so comparing them identifies which activity consumed the most time during the unusually long run. This is the direct way to locate the bottleneck at activity granularity.

Why this answer

Pipeline run details in the Monitoring hub decompose a run into its activities with per-activity status and duration. Drilling into the unusually long run and comparing those durations isolates the activity that consumed the most time, which is exactly the diagnostic the engineer requested and the prerequisite for any remediation.

Exam trap

The trap here is treating capacity-level consumption reporting as if it provided per-activity run diagnostics, when activity duration detail comes from the run view in the Monitoring hub.

11
MCQhard

Refer to the exhibit. The Spark job completed but took 45 minutes for a 100 million row dataset. The shuffle write size is significantly high. What should you investigate to improve performance?

A.Reduce the number of partitions.
B.Investigate join strategies and broadcast settings.
C.Enable auto-vacuuming on the table.
D.Increase the disk capacity of the Lakehouse.
AnswerB

High shuffle write sizes are a classic symptom of inefficient join operations. By analyzing the execution plan in the Spark UI, you can determine if a broadcast join would be more appropriate, which would allow the engine to avoid the expensive shuffle phase entirely for smaller lookup tables.

Why this answer

High shuffle write size in a Spark job typically indicates expensive wide transformations such as joins or group-bys that move data across the network. Investigating join strategies — especially whether broadcast joins can be used for small dimension tables — and tuning the broadcast threshold or join hints directly reduces shuffle volume and runtime.

Exam trap

DP-700 often tests the reflex to blame partitioning or storage when shuffle is high, but the correct diagnostic path for large shuffle writes is join strategy and broadcast settings — candidates must distinguish shuffle I/O from partition count and storage capacity.

How to eliminate wrong answers

Option A is wrong because reducing the number of partitions generally increases per-partition data size and can worsen shuffle and skew; the goal is usually to right-size partitions (e.g., via AQE coalescing), not blindly reduce them. Option C is wrong because auto-vacuuming (Delta Lake VACUUM) removes old files to reclaim storage and does not affect shuffle write size or join performance. Option D is wrong because increasing Lakehouse disk capacity addresses storage limits, not shuffle I/O or join execution — shuffle is a memory/network/disk-spill concern, not a capacity quota issue.

12
Multi-Selectmedium

You are optimizing an analytics solution in Microsoft Fabric. A Data Warehouse serves both a high-volume nightly ELT load and interactive analyst queries during the day. You observe that daytime queries are slow and that load jobs overlap with query activity. You need to reduce contention between the two workloads. (Choose two.)

Select 2 answers
A.Route the interactive analyst queries through a separate workspace with its own capacity.
B.Use separate schemas to define staging and serving tables so load operations target staging objects.
C.Create a columnstore index on every table in the warehouse.
D.Schedule the nightly ELT load to complete before the daytime query window begins.
E.Enable result-set caching on the warehouse to reuse prior query output.
AnswersA, D

Placing interactive queries on a distinct capacity physically isolates compute from the ELT load, so a heavy load cannot starve analyst queries. This is workload isolation at the capacity level, which directly addresses contention. It costs additional capacity, but it reliably prevents the nightly load from degrading the daytime interactive experience, satisfying the isolation requirement.

Why this answer

Contention between a heavy ELT load and interactive queries is best addressed by isolating them in time or in compute. Finishing the load before the query window separates the workloads temporally, while moving analyst queries to a capacity of their own separates them physically. Schema separation, result caching, and broad columnstore indexing change organization or storage behavior but leave the two workloads competing for the same resources.

Exam trap

The trap here is treating schema separation or caching as workload isolation, when only scheduling the load away from query hours or giving queries their own capacity actually separates the competing workloads.

13
MCQhard

Refer to the exhibit. A data engineer is troubleshooting a Spark job that frequently fails with 'Out of Memory' (OOM) errors despite the configuration shown. The job processes a 500GB dataset with many wide transformations. What is the most effective configuration change to resolve the OOM errors?

A.Set dynamicAllocation to false and use 50 static executors.
B.Increase the executorCores to 8 to allow more tasks to run.
C.Increase executorMemory and decrease executorCores per executor.
D.Set vorderEnabled to false to reduce the memory overhead.
AnswerC

By increasing memory and decreasing cores, you provide more memory to each individual task running on that executor. This is a standard approach for handling wide transformations or large shuffles where each task requires a significant amount of memory to store intermediate data structures without spilling to disk or crashing.

Why this answer

Increasing executorMemory while decreasing executorCores reduces the number of concurrent tasks per executor, which lowers the per-executor memory contention caused by many simultaneous tasks. With wide transformations, each task holds shuffle buffers, sort buffers, and cached partitions; fewer concurrent tasks per JVM means each task gets a larger share of the heap, preventing OOM. This is the standard Spark tuning pattern for memory-bound jobs with wide dependencies.

Exam trap

The trap here is assuming 'more cores = more parallelism = better', when in Spark more cores per executor actually increases concurrent memory consumption and is a classic cause of executor OOM.

How to eliminate wrong answers

Option A is wrong because disabling dynamic allocation and pinning 50 static executors does not address per-executor memory pressure and can actually waste cluster resources or worsen skew. Option B is wrong because increasing executorCores to 8 increases concurrent task count per JVM, multiplying memory pressure from shuffle/sort buffers and making OOM more likely. Option D is wrong because 'vorderEnabled' is not a real Spark configuration parameter (likely a distractor resembling 'spark.sql.execution.sortBeforeRepartition' or similar), so toggling it cannot resolve OOM.

14
MCQeasy

You are responsible for monitoring the health of a Microsoft Fabric capacity. You need to receive proactive notifications when the capacity utilization exceeds 80% for more than 15 minutes. Which Fabric feature should you use to set up these alerts?

A.Monitor hub
B.Azure Monitor
C.Capacity Metrics app
D.Data Activator
AnswerD

Data Activator in Microsoft Fabric allows you to set up alerts based on conditions in real-time data streams. You can define a rule that triggers when capacity utilization exceeds 80% for a sustained period, sending notifications via email or Teams. This directly meets the requirement for proactive alerts.

Why this answer

Data Activator is the native Fabric component for setting up alerts based on conditions in data streams. It can monitor capacity metrics and trigger notifications when thresholds are breached. The Capacity Metrics app and Monitor hub are for viewing data, and Azure Monitor is external and requires more setup.

Exam trap

The trap here is confusing the Capacity Metrics app, which shows data, with an alerting mechanism, leading to the selection of a monitoring tool instead of an alerting tool.

15
MCQmedium

You are monitoring a Spark notebook in Microsoft Fabric that is taking longer than expected to process a large dataset. You notice that a single executor is processing significantly more data than others. Which tool or feature should you use to identify the specific partition causing this data skew?

A.Fabric Monitoring Hub
B.The Spark Advisor and Spark UI
C.The Capacity Metrics App
D.The OneLake Explorer
AnswerB

The Spark UI provides a detailed breakdown of stages, tasks, and executors, allowing you to see the exact bytes read by each partition. Combined with Spark Advisor, it offers actionable recommendations to resolve skew, such as using salting techniques or repartitioning the data to ensure a more even distribution of work.

Why this answer

The Spark Advisor and Spark UI provide detailed insights into Spark job execution, including stage and task-level metrics. The Spark UI shows task distribution and can highlight skew by revealing tasks that process significantly more data than others. Spark Advisor automatically detects skew and provides recommendations.

Exam trap

DP-700 often tests the confusion between high-level monitoring tools (Monitoring Hub, Capacity Metrics) and the detailed Spark UI/Advisor needed for diagnosing data skew.

How to eliminate wrong answers

Option A is wrong because the Fabric Monitoring Hub provides high-level job status and duration, but not the granular task-level details needed to identify data skew. Option C is wrong because the Capacity Metrics App monitors capacity utilization, not individual Spark job internals. Option D is wrong because OneLake Explorer is for browsing and managing data in OneLake, not for debugging Spark jobs.

16
MCQmedium

You are the data engineer for a Microsoft Fabric workspace that ingests streaming events into an Eventstream and lands them in a Lakehouse Delta table via a pipeline. Business users report that dashboards built on the Silver layer show stale data for several hours each morning, even though the pipeline run history shows 'Succeeded'. You need to determine why the downstream semantic model is not reflecting new rows. Which action should you take first to diagnose the issue?

A.Enable automatic schema drift handling on the Lakehouse Delta table so that new event columns are merged during ingestion.
B.Open the Monitoring hub for the workspace and inspect the refresh history of the affected semantic model to compare its last refresh time against the pipeline completion time.
C.Reconfigure the Eventstream to use a larger retention period so that no events are dropped before the pipeline reads them.
D.Increase the Fabric capacity SKU to the next tier so that background operations stop being throttled during the morning window.
AnswerB

The Monitoring hub aggregates activity across pipelines, semantic model refreshes, and Spark jobs, so comparing the semantic model refresh timestamp against pipeline completion reveals whether the model simply has not refreshed after the new rows landed. This is the fastest first diagnostic because it distinguishes a refresh-timing problem from a data-landing problem without changing any configuration.

Why this answer

A pipeline that reports success only guarantees that data landed in the Lakehouse; it says nothing about whether the semantic model was refreshed afterward. The Monitoring hub surfaces both pipeline runs and semantic model refreshes in one timeline, letting you confirm whether the model refresh simply lags the ingestion. Adjusting capacity, Eventstream retention, or schema drift all target unrelated causes and would not explain a clean pipeline success with stale reporting.

Exam trap

The trap here is assuming that a successful pipeline run also refreshes the semantic model, when ingestion and model refresh are separate operations with independent schedules.

17
MCQmedium

You are monitoring a Microsoft Fabric Lakehouse. You notice that queries against a specific Delta table are performing slowly despite the table having a small data size. You need to identify the root cause of the performance degradation. What should you examine first?

A.The number of partitions in the table metadata.
B.The total number of files within the table directory.
C.The CPU utilization of the Spark pool.
D.The storage account throughput limits.
AnswerB

Delta tables perform best when data is stored in optimally sized files. If a table contains thousands of small files, the query engine spends excessive time on metadata listing and initialization. Monitoring file count helps identify when the table requires a compaction process to merge these small files.

Why this answer

In a Microsoft Fabric Lakehouse, Delta tables store data as Parquet files in a directory, and query performance is heavily influenced by file count rather than total data size. A small table fragmented into thousands of tiny files causes excessive metadata overhead, poor parallelism, and slow file listing during query planning. Examining the number of files in the table directory is the fastest way to confirm this 'small files problem.'

Exam trap

DP-700 often tests whether candidates blame compute (Spark pool CPU) or storage throughput for slow queries when the actual root cause is Delta table file fragmentation — the trap is overlooking the small-files problem.

How to eliminate wrong answers

Option A is wrong because partition count matters for pruning, but a small table with slow queries is more likely suffering from file fragmentation than from partition design — and partition metadata alone won't reveal the small-file issue. Option C is wrong because Spark pool CPU utilization would show resource contention across all workloads, not a table-specific slowdown; a small table with slow queries is unlikely to be CPU-bound. Option D is wrong because storage account throughput limits would affect all tables and workloads, not just one small Delta table, and would typically manifest as throttling errors rather than consistent slowness.

18
MCQeasy

Which tool in Fabric provides the most detailed view of the billable capacity usage for your tenant?

A.The Microsoft Cost Management portal
B.The Fabric Capacity Metrics app
C.Azure Advisor
D.The Fabric Admin Portal
AnswerB

The Capacity Metrics app is the standard Fabric tool for viewing detailed consumption data. It provides the necessary insights to correlate actual usage against your purchased capacity units, helping you understand your spend and identify opportunities for optimization or right-sizing your SKU.

Why this answer

The Fabric Capacity Metrics app is a Power BI template app provided by Microsoft that gives the most detailed, granular view of billable capacity usage across the tenant. It breaks down consumption by workspace, item, and operation, showing CU (Capacity Unit) usage, throttling, and overage. This is the authoritative tool for capacity monitoring and chargeback.

Exam trap

The trap is confusing cost visibility (Cost Management) with capacity telemetry (Capacity Metrics app) — candidates often pick the billing portal because the question mentions 'billable' usage.

How to eliminate wrong answers

Option A is wrong because the Microsoft Cost Management portal shows Azure billing costs but lacks the granular Fabric workload-level detail (per-item CU consumption) needed for capacity analysis. Option C is wrong because Azure Advisor provides optimization recommendations, not detailed usage metrics. Option D is wrong because the Fabric Admin Portal offers tenant-level settings and some usage metrics, but not the deep, billable CU breakdown that the Capacity Metrics app provides.

19
MCQmedium

A Fabric pipeline that ingests data from an on-premises SQL Server has been intermittently failing. You need to see how many times it has failed recently and the duration of each run, and you want a single place that shows run history for the pipeline. Where should you look?

A.The Fabric Capacity Metrics app, using the Timepoint page to read the failure count and duration of the pipeline.
B.The Monitoring hub, filtered to the pipeline activity, showing run status, start time, and duration for each execution.
C.The pipeline's own Run history, opened from the pipeline item, listing each run with its status and timing.
D.The on-premises data gateway logs, which record each pipeline execution and its outcome.
AnswerB

The Monitoring hub provides a consolidated view of Fabric activities, including pipeline runs, with status, start and end times, and duration, and it can be filtered by item or activity type. That makes it the single place to see recent failures and per-run durations for the pipeline, matching the requirement exactly.

Why this answer

The Monitoring hub is Fabric's consolidated activity view and can be filtered to show pipeline runs with status, start time, and duration, so it answers both how many failures occurred and how long each run took. Item-level run history is accurate but scoped to one pipeline, while capacity metrics and gateway logs describe resource usage and connectivity rather than pipeline run outcomes.

Exam trap

The trap here is assuming that capacity-level or gateway-level telemetry contains pipeline run status and duration, when run outcomes live in the activity monitoring surfaces.

20
MCQhard

You are auditing data access in a Microsoft Fabric workspace and notice unexpected 'Access Denied' errors in the OneLake logs for a service principal that should have read-only access. The service principal has the 'Viewer' role in the workspace. What is the most likely cause?

A.The service principal needs the 'Contributor' role to read data.
B.OneLake data access control (preview) is restricting specific folders.
C.The service principal has exceeded its OneLake throughput quota.
D.OneLake does not support service principal authentication.
AnswerB

When OneLake data access control is enabled, it allows for more granular security than workspace roles. Even with a 'Viewer' role, if the specific path or folder in the Lakehouse has been restricted via these granular permissions, the service principal will receive an 'Access Denied' error when attempting to read.

Why this answer

OneLake data access control (preview) allows workspace admins to restrict access to specific folders within a lakehouse, overriding workspace-level roles like Viewer. A service principal with the Viewer role can read the workspace, but if OneLake data access control is configured to deny access to certain folders, the service principal will receive 'Access Denied' errors when attempting to read those folders. This is the most likely cause given the service principal already has read-only workspace access.

Exam trap

DP-700 often tests the interaction between workspace roles and OneLake folder-level permissions — candidates assume Viewer role guarantees read access to all data, but OneLake data access control can override workspace roles and deny specific folders, producing 'Access Denied' despite adequate workspace-level permissions.

How to eliminate wrong answers

Option A is wrong because the Viewer role already grants read-only access to workspace items — Contributor is not required for reading and would grant excessive privileges (write/delete), which is not the issue. Option C is wrong because OneLake does not enforce a per-service-principal throughput quota that would manifest as 'Access Denied' — throttling would produce different errors (429/timeout), not authorization failures. Option D is wrong because OneLake fully supports service principal authentication via Microsoft Entra ID; service principals are a supported and recommended identity type for automated access.

21
MCQhard

You are monitoring a Microsoft Fabric workspace where a Data Pipeline loads data into a Lakehouse and a downstream notebook transforms it. A business user reports that a Power BI report shows yesterday's data even though the pipeline run for today completed successfully. You need to determine why the report is stale despite the successful pipeline run, and you must identify the earliest point in the chain to investigate. What should you do first?

A.Increase the pipeline's concurrency limit so the notebook and pipeline run in parallel.
B.Open the Capacity Metrics app and check whether the capacity was throttled during the pipeline window.
C.Recreate the semantic model and republish the Power BI report to force a refresh.
D.Inspect the notebook run details in the Monitoring hub to verify the transformation wrote the expected partition and row counts.
AnswerD

The pipeline succeeded, but success at the orchestration layer does not guarantee the notebook produced the correct output. Reviewing the notebook run in the Monitoring hub shows its status, duration, and output details, letting you confirm whether the transformation actually wrote today's partition and expected rows. This is the earliest downstream step that could leave the report stale.

Why this answer

A successful pipeline run only confirms orchestration completed, not that downstream transformations produced correct data. The first place to verify is the notebook run, since it writes the transformed output that the report consumes. Confirming the expected partition and row counts there isolates whether the break is in transformation or later in the semantic model refresh.

Exam trap

The trap here is equating a successful pipeline run with correct end-to-end data, when orchestration success says nothing about whether downstream transformations wrote the right output.

22
MCQmedium

You are responsible for a Microsoft Fabric Data Warehouse that experiences performance degradation during peak hours. You suspect that a small number of queries are consuming a disproportionate amount of resources. You need to identify the top resource-consuming queries over the past week. What should you do?

A.Run dynamic management views (DMVs) to capture currently executing queries and their resource usage.
B.Configure the Capacity Metrics app to send alerts when resource usage exceeds a threshold.
C.Use SQL Server Profiler connected to the Fabric Data Warehouse.
D.Use the Query Activity view in the Fabric Data Warehouse to analyze historical query performance.
AnswerD

The Query Activity view in Fabric Data Warehouse provides historical query execution data, including resource consumption metrics such as CPU and duration. By filtering for the past week and sorting by resource usage, you can identify the top-consuming queries and take corrective action, such as optimizing or scheduling them during off-peak hours.

Why this answer

The Query Activity view in Fabric Data Warehouse retains historical query data for up to 30 days, including resource metrics. By analyzing this view over the past week, you can pinpoint the queries with the highest resource consumption. DMVs only show current activity, the Capacity Metrics app aggregates at the capacity level, and SQL Server Profiler is not available for Fabric, making the Query Activity view the correct tool.

Exam trap

The trap here is assuming that on-premises tools like SQL Server Profiler or DMVs can be used in Fabric for historical analysis.

23
MCQhard

You are a data engineer monitoring a Microsoft Fabric Lakehouse. You notice that a Delta table is experiencing performance degradation over time due to a large number of small files. You need to optimize the table to improve read performance. What should you do?

A.Run the OPTIMIZE command on the Delta table to compact small files.
B.Increase the retention period for Delta log files to improve read performance.
C.Use the VACUUM command to remove old files and free up space.
D.Convert the Delta table to Parquet format to reduce overhead.
AnswerA

The OPTIMIZE command in Delta Lake compacts small files into larger ones, reducing the number of files and improving read performance. In Microsoft Fabric Lakehouse, you can run OPTIMIZE on a Delta table to address the small file problem. This operation is essential for maintaining query performance as data is frequently appended, which can create many small files.

Why this answer

The OPTIMIZE command in Delta Lake compacts small files into larger ones, which reduces the number of files that need to be read and improves query performance. This is the standard remedy for the small file problem in Delta tables. VACUUM only removes old files and does not compact, while converting to Parquet or adjusting log retention do not address the issue.

Exam trap

The trap here is thinking that VACUUM can solve small file issues; VACUUM only removes unreferenced files, while OPTIMIZE is required to compact them.

24
MCQmedium

You manage a Microsoft Fabric workspace containing a Data Warehouse. Analysts report that a stored procedure loading the fact_sales table takes progressively longer each night, and the duration increase correlates with the growing number of rows in that table. You need to identify the root cause using built-in Fabric tooling before making any changes. Which action should you take first?

A.Increase the capacity SKU assigned to the workspace so the stored procedure receives more compute resources during the nightly load.
B.Rebuild all column statistics on the fact_sales table and rerun the procedure to see whether the duration returns to normal.
C.Open the query insights view for the Data Warehouse and review the historical query plan and runtime statistics for the stored procedure.
D.Enable verbose logging in the pipeline that calls the stored procedure and inspect the activity run history for error messages.
AnswerC

Query insights in a Fabric Data Warehouse retains historical query text, durations, and execution plans, so you can see exactly which operator in the stored procedure degrades as row counts grow. This is the correct first diagnostic step because it produces evidence before any tuning change is made, and it works without redeploying or modifying the procedure.

Why this answer

Query insights is the Fabric Data Warehouse feature that preserves historical execution data, including query text, duration trends, and plans, which is precisely what is needed to explain a gradual slowdown tied to table growth. Capturing that evidence first keeps the investigation grounded and lets you verify any later tuning change against a real baseline rather than guessing at capacity or statistics.

Exam trap

The trap here is assuming that a slowing nightly load must be a resource shortage and jumping to capacity scaling instead of capturing the historical plan that would reveal the actual operator-level cause.

25
Multi-Selectmedium

You manage a Fabric workspace with multiple Lakehouses and Data Warehouses. You need to identify which operations are consuming the most Fabric capacity and causing throttling during business hours. (Choose two.)

Select 2 answers
A.Run sp_who2 on the Fabric Data Warehouse to see active sessions and resource usage.
B.Analyze the capacity's utilization metrics to identify the top consumers during peak hours.
C.Use the Microsoft Fabric Capacity Metrics app to view the utilization and throttling of the capacity.
D.Review the Query activity in the Monitoring hub for each item in the workspace.
E.Enable verbose logging on all notebooks and pipelines in the workspace.
AnswersB, C

Analyzing the capacity's utilization metrics is exactly how you find the operations consuming the most capacity. The Capacity Metrics app provides these metrics, breaking down usage by item and operation type. This lets you correlate heavy usage with business-hour throttling and is a correct approach for the stated goal.

Why this answer

To find which operations consume the most capacity and cause throttling, you need capacity-level telemetry. The Microsoft Fabric Capacity Metrics app provides utilization and throttling insights, and analyzing those utilization metrics identifies the top consumers during peak periods. Monitoring hub, SQL session procedures, and verbose logging do not deliver that aggregated capacity view.

Exam trap

The trap here is choosing per-item monitoring tools like the Monitoring hub or verbose logs, which show run status but not aggregated capacity consumption or throttling.

26
MCQeasy

Which feature in Microsoft Fabric allows you to track the movement and transformation of data from source to destination?

A.Data Pipeline
B.Data Lineage
C.OneLake
D.Notebook
AnswerB

Data Lineage is the feature in Fabric that maps the relationship between various data artifacts. It provides a visual graph showing how data flows from sources into Lakehouses, warehouses, and ultimately into semantic models and reports, which is essential for impact analysis.

Why this answer

Data Lineage in Microsoft Fabric provides a visual, end-to-end view of how data moves and transforms from source systems through pipelines, notebooks, and dataflows into destinations like lakehouses and warehouses. It records the relationships between datasets and their upstream/downstream dependencies, which is exactly what tracking movement and transformation requires.

Exam trap

The trap is confusing the tool that performs data movement (Data Pipeline) with the feature that documents it (Data Lineage); candidates pick the pipeline because it 'moves data' but the question asks about tracking.

How to eliminate wrong answers

Option A is wrong because a Data Pipeline orchestrates and executes data movement and transformation, but it does not itself provide the lineage view that traces data across artifacts. Option C is wrong because OneLake is the unified storage layer for Fabric, not a tracking or lineage feature. Option D is wrong because a Notebook is a code authoring surface for running Spark/Python; it executes transformations but does not visualize or track lineage.

27
MCQhard

You are monitoring a Fabric Data Pipeline that copies data from an on-premises SQL Server into a Lakehouse. The pipeline occasionally fails with a timeout error on the source connection, but when you manually rerun the same activity it succeeds. You need to make the pipeline more resilient to these transient source timeouts without masking permanent failures. What should you configure on the Copy activity?

A.Add a Wait activity before the Copy activity with a long delay.
B.Increase the DIU (Data Integration Unit) setting for the Copy activity.
C.Set the activity's retry policy to a fixed interval with a small number of retries.
D.Enable the 'Logging' diagnostic setting on the pipeline and alert on failures.
AnswerC

A retry policy on the Copy activity is the intended mechanism for transient failures. Configuring a small number of retries with a fixed interval allows the pipeline to automatically recover from intermittent source timeouts, while the limited retry count prevents an infinite loop when the failure is permanent. This directly matches the requirement to be resilient without masking real errors.

Why this answer

Transient source timeouts are best handled by a retry policy on the Copy activity. A small number of retries with a fixed interval lets the pipeline recover automatically from intermittent failures, while the capped retry count ensures that persistent problems still surface as failures. Other settings affect throughput or observability, not automatic recovery.

Exam trap

The trap here is confusing observability features like logging and alerting with actual resilience mechanisms, or assuming more compute prevents connection timeouts.

28
MCQmedium

You are troubleshooting a slow-running Power BI report connected to a Fabric Semantic Model. What is the most effective way to identify the bottleneck?

A.Use the Power BI Performance Analyzer
B.View the workspace storage metrics
C.Check the Fabric capacity SKU settings
D.Rebuild the semantic model from scratch
AnswerA

Performance Analyzer records the time taken for each visual to refresh, including DAX query execution and visual rendering. It provides a detailed breakdown of where time is spent, making it the most effective tool for identifying bottlenecks within a report's specific components.

Why this answer

Power BI Performance Analyzer captures detailed timing data for every visual, DAX query, and rendering operation in the report, showing exactly which visual or query consumes the most time. This makes it the only option that directly pinpoints the bottleneck at the report-to-semantic-model interaction level. It records events such as 'Query Start/End' and 'Visual Rendering' with millisecond precision, enabling targeted optimization.

Exam trap

DP-700 often tests the misconception that capacity or storage metrics can diagnose report performance, when only Performance Analyzer provides per-visual query timing.

How to eliminate wrong answers

Option B is wrong because workspace storage metrics only show capacity consumption and dataset sizes, not query execution times or visual rendering performance. Option C is wrong because Fabric capacity SKU settings indicate the compute tier and throttling limits but do not reveal which specific report operation is slow. Option D is wrong because rebuilding the semantic model is a drastic, time-consuming action that does not diagnose the bottleneck and may not resolve it.

29
MCQeasy

You are a data engineer responsible for monitoring a Microsoft Fabric Data Warehouse. You need to identify long-running queries that are consuming significant resources and causing performance degradation. Which feature in Microsoft Fabric should you use to analyze query performance and identify bottlenecks?

A.Microsoft Fabric Capacity Metrics app.
B.Azure Monitor.
C.SQL Server Profiler.
D.Dynamic Management Views (DMVs) in the Data Warehouse.
AnswerD

DMVs in Microsoft Fabric Data Warehouse provide detailed information about query execution, resource usage, and performance metrics. They allow you to identify long-running queries, waits, and bottlenecks, which is exactly what is needed to analyze and optimize query performance in this scenario.

Why this answer

Dynamic Management Views (DMVs) are the primary tool for analyzing query performance in Microsoft Fabric Data Warehouse. They provide insights into query execution plans, resource consumption, and wait statistics, enabling you to pinpoint long-running queries and optimize them. The Capacity Metrics app and Azure Monitor are more focused on capacity-level monitoring, while SQL Server Profiler is not available in Fabric.

Exam trap

The trap here is confusing capacity-level monitoring tools with query-level diagnostics, which are best handled by DMVs in the Data Warehouse.

30
MCQmedium

You have a Microsoft Fabric Lakehouse that stores large Delta tables. You notice that queries against a specific table are slow, and you want to reduce the amount of data scanned during query execution. The table is frequently filtered on a column named TransactionDate. Which optimization technique should you implement to improve query performance?

A.Increase the size of the Spark pool.
B.Convert the table to Parquet format.
C.Partition the table by TransactionDate.
D.Enable caching for the Lakehouse.
AnswerC

Partitioning the Delta table by TransactionDate organizes data into separate folders based on the date. When queries filter on TransactionDate, the engine can skip irrelevant partitions, drastically reducing the amount of data scanned. This directly addresses the slow queries caused by full table scans.

Why this answer

Partitioning by TransactionDate allows the query engine to perform partition pruning, reading only the partitions that match the filter. This significantly reduces I/O and speeds up queries. Other options either do not address data volume or are already inherent to Delta tables.

Exam trap

The trap here is thinking that increasing compute resources or changing file format will reduce data scanned, when the real issue is lack of partition pruning.

31
MCQhard

A Fabric workspace runs a Data Pipeline that copies data from an on-premises SQL Server via a gateway, then a notebook transforms it. Loads that used to finish in 20 minutes now take over 2 hours, and the Monitoring hub shows the copy activity itself as the long-running step while the notebook remains quick. You must determine why the copy activity slowed down without changing the pipeline design. Which action should you take first?

A.Open the copy activity's detailed run in the Monitoring hub and compare its read and write durations, throughput and data volume against earlier successful runs.
B.Restart the on-premises data gateway service and clear its cache, because gateway caching is the most common cause of slow copy performance.
C.Enable staging on the copy activity so data is written to a temporary location in OneLake before being loaded into the destination.
D.Increase the degree of copy parallelism on the copy activity and rerun the pipeline to see whether throughput improves.
AnswerA

The Monitoring hub exposes per-activity run details for pipelines, including copy duration, rows and bytes transferred, throughput and the split between reading from the source and writing to the sink. Comparing those figures against earlier runs shows whether the source read slowed, the sink write slowed, or the volume changed, which isolates the cause without altering the pipeline and directly satisfies the requirement to determine why the copy slowed.

Why this answer

The Monitoring hub holds per-activity copy run details that break duration into source read and sink write components, along with rows, bytes and throughput. Comparing the degraded runs with earlier healthy ones shows whether the source is returning data more slowly, the sink has become the constraint, or the volume has grown, all without modifying the pipeline. Remediation options such as parallelism, staging or gateway restarts should follow only after the bottleneck is identified.

Exam trap

The trap here is jumping to a configuration change such as copy parallelism or staging, when the Monitoring hub already contains the per-activity evidence needed to locate the bottleneck.

32
MCQeasy

You are monitoring a Microsoft Fabric workspace that contains several Data Pipelines, notebooks, and a Data Warehouse. Users report that a scheduled pipeline intermittently fails, but the failure message does not indicate which activity failed. You need to identify the exact activity and the error details for each failed run, and you want to review this information centrally without opening each pipeline individually. What should you use?

A.The workspace lineage view, tracing the upstream dependencies of the pipeline.
B.The Lakehouse table maintenance logs stored in the OneLake diagnostics folder.
C.The Fabric Capacity Metrics app, reviewing the timepoints for throttling.
D.The Monitoring hub, filtering by item type and status to inspect activity-level run details.
AnswerD

The Monitoring hub in Microsoft Fabric provides a centralized view of run history across items such as pipelines, notebooks, and dataflows. It shows per-activity status and error messages, so you can identify exactly which activity failed and read its error details without opening each pipeline. Filtering by item type and status narrows the list to the failing runs you need to investigate.

Why this answer

Diagnosing an intermittent pipeline failure at the activity level requires per-run execution history with error messages. The Monitoring hub aggregates runs across Fabric items and exposes activity-level status and errors, which lets you pinpoint the failing activity centrally. Capacity metrics, lineage, and maintenance logs describe resource usage, relationships, or table operations rather than activity execution outcomes.

Exam trap

The trap here is reaching for the Capacity Metrics app because it is the familiar monitoring surface, when activity-level error diagnostics live in the Monitoring hub.

33
MCQhard

You are optimizing a Spark job in a Fabric notebook that joins a large fact table with a small dimension table. The join operation is causing high memory usage and slow performance. What technique should you implement?

A.Increase the number of partitions for the large table.
B.Use a broadcast join hint.
C.Force a cross join.
D.Enable dynamic partition pruning.
AnswerB

Broadcast join hints instruct the Spark optimizer to send the smaller table to all worker nodes. This eliminates the shuffle phase, allowing the join to happen locally on each node, which is the most efficient way to handle fact-dimension joins when one table is relatively small.

Why this answer

A broadcast join hint (e.g., /*+ BROADCAST(dim_table) */ or broadcast(dim_df)) tells Spark to replicate the small dimension table to every executor and perform a map-side join, eliminating the expensive shuffle of the large fact table. This dramatically reduces network I/O and memory pressure because the large table's partitions are processed locally against the in-memory broadcast copy. Spark's default threshold (spark.sql.autoBroadcastJoinThreshold, typically 10 MB) may not trigger automatically, so an explicit hint is warranted.

Exam trap

The trap here is confusing partition-count tuning with join-strategy optimization; candidates often pick 'increase partitions' because it sounds like a performance fix, but the exam expects recognition that broadcast joins eliminate the shuffle entirely for skewed large-small joins.

How to eliminate wrong answers

Option A is wrong because increasing partitions on the large table adds more shuffle tasks and overhead without addressing the root cause — the shuffle itself — and can worsen small-file/task-scheduling overhead. Option C is wrong because a cross join produces a Cartesian product (N×M rows), which explodes data volume and memory usage catastrophically rather than reducing it. Option D is wrong because dynamic partition pruning only helps when filtering on partition columns of the probe side at runtime; it does not eliminate the shuffle required for a standard sort-merge join of a large fact with a small dimension.

34
MCQmedium

Refer to the exhibit. You are reviewing pipeline logs and notice the 429 error. What is the most appropriate long-term action to prevent this?

A.Increase the Fabric capacity SKU immediately
B.Configure retry policies with exponential backoff
C.Analyze query patterns and optimize resource-heavy jobs
D.Disable monitoring to reduce overhead
AnswerC

Optimizing resource-intensive jobs is the most sustainable way to resolve 429 errors. By improving query efficiency, reducing data volume, or staggering pipeline executions, you ensure that the current capacity can handle the workload without hitting the hard compute limits defined by the service.

Why this answer

A 429 error indicates that the request was throttled because the capacity reached its compute limit. While immediate scaling might fix the current issue, long-term stability requires identifying the root cause of the spike in resource usage. Optimizing the code, adjusting concurrency, or smoothing out scheduled job times ensures the capacity operates within its allocated limits without requiring constant upgrades.

Exam trap

Candidates frequently select immediate capacity SKU upgrades as the primary answer, missing the prompt's focus on long-term prevention through code and query optimization.

35
MCQmedium

You have a Fabric notebook that uses multiple Spark libraries. You notice that the startup time for your notebook is very slow. What is the best way to optimize this?

A.Install libraries using %pip at the top of the notebook.
B.Use a pre-configured Fabric Environment.
C.Store all libraries in a Lakehouse folder.
D.Increase the Spark cluster size.
AnswerB

Fabric Environments are pre-built configurations that include libraries and settings, allowing them to be loaded efficiently during session initialization. Using them avoids the latency of runtime installations and ensures all notebooks in the workspace share a consistent, performant runtime environment.

Why this answer

A pre-configured Fabric Environment packages the required Spark libraries and their dependencies into a reusable artifact that is attached to the notebook session at startup, so the libraries are already resolved and cached when the session begins. This eliminates the repeated download, dependency resolution, and installation overhead that occurs when libraries are installed at runtime. Because the environment is pre-built, session start time is significantly reduced compared to ad-hoc installation methods.

Exam trap

DP-700 often tests the distinction between runtime library installation (%pip, %conda) and pre-provisioned environments, tricking candidates into choosing the seemingly convenient %pip approach when the question explicitly asks about startup optimization.

How to eliminate wrong answers

Option A is wrong because %pip installs libraries at runtime on every session start, forcing Spark to download and resolve dependencies each time, which is precisely the cause of slow startup. Option C is wrong because storing libraries in a Lakehouse folder does not make them available to the Spark session — they still need to be installed or referenced, and Lakehouse storage is not a library resolution mechanism. Option D is wrong because increasing cluster size adds compute capacity but does not address the library installation and dependency resolution overhead that dominates startup time.

36
Multi-Selectmedium

You are monitoring a Dataflow Gen2 refresh that is failing intermittently. You need to identify if the failure is caused by a data type mismatch or a timeout from the source system. Which TWO actions will help you find the specific error details? (Choose two.)

Select 2 answers
A.Check the 'Refresh history' in the Dataflow settings.
B.Enable 'Stage query' for all transformations.
C.View the 'DataflowRefreshHistory' table in the Lakehouse.
D.Examine the 'On-premises data gateway' logs if applicable.
E.Use the 'Performance Analyzer' in Power BI Desktop.
AnswersA, D

The Refresh History page provides a list of all past refresh attempts, their duration, and a summary of success or failure. For failed runs, it often provides an initial error message or a link to download a more detailed error log that contains the underlying Power Query exception.

Why this answer

Option A is correct because the 'Refresh history' in the Dataflow settings records each refresh run with its status and error message, which is where a data type mismatch or source timeout error surfaces for Dataflow Gen2. Option D is correct because if the source is reached through an on-premises data gateway, the gateway logs capture connection, timeout, and query execution failures that the cloud-side refresh history may only summarize. Options B, C, and E do not belong: 'Stage query' is a Power Query authoring option that materializes intermediate steps and does not expose refresh error details; a 'DataflowRefreshHistory' table in the Lakehouse is not the standard artifact for retrieving Dataflow Gen2 refresh errors; and Performance Analyzer in Power BI Desktop measures report/visual performance, not Dataflow Gen2 refresh failures.

Exam trap

DP-700 often tests the confusion between Dataflow refresh history and other monitoring tools like Performance Analyzer or Lakehouse tables, leading candidates to choose incorrect locations for error details.

37
MCQhard

You are monitoring a Microsoft Fabric Data Warehouse that experiences high concurrency during business hours. Users report that some queries are slow, and you suspect that certain queries are waiting on resources. You need to identify which queries are currently running and which are blocked, and you want to see the blocking chain. Which dynamic management view (DMV) should you query?

A.sys.dm_exec_sessions
B.sys.dm_exec_query_stats
C.sys.dm_db_index_usage_stats
D.sys.dm_exec_requests
AnswerD

sys.dm_exec_requests returns information about each request currently executing in the Data Warehouse, including the blocking_session_id column that identifies blocking chains. By querying this DMV, you can see which sessions are blocked and by whom, directly addressing the need to identify blocking queries and their relationships.

Why this answer

sys.dm_exec_requests is the correct DMV because it includes the blocking_session_id column, which reveals blocking chains. It also shows the current SQL text, wait type, and wait time for each request. The other DMVs provide session, query plan, or index statistics but do not capture active blocking relationships.

Exam trap

The trap here is assuming that session-level or aggregate performance DMVs can reveal blocking, when only the requests DMV includes the blocking_session_id column.

38
MCQmedium

You are a data engineer for a retail company using Microsoft Fabric. A Data Pipeline that loads sales data into a Lakehouse is scheduled to run every 15 minutes. You receive alerts that the pipeline is intermittently failing with a 'RequestThrottled' error. You need to determine the root cause. You open the Microsoft Fabric Capacity Metrics app and notice that the 'Background' operations are consistently exceeding the capacity limit. What should you do to resolve the issue?

A.Move the pipeline to a different workspace with a higher capacity.
B.Enable autoscale for the capacity to automatically add resources during peak times.
C.Optimize the pipeline to reduce the number of background operations, such as by batching or reducing the frequency of the pipeline runs.
D.Increase the capacity size to the next SKU.
AnswerC

The pipeline is running every 15 minutes, generating many background operations that exceed the capacity limit. By optimizing the pipeline to batch data or reducing its frequency, you lower the number of background operations, thereby reducing the chance of hitting the capacity limit and avoiding throttling.

Why this answer

The 'RequestThrottled' error occurs when a Fabric capacity exceeds its limits for background operations. The pipeline's high frequency (every 15 minutes) generates many background operations, causing the capacity to throttle. Reducing the frequency or batching operations lowers the load, preventing throttling.

Increasing capacity size or moving workspaces may provide temporary relief but does not address the root cause of excessive operations.

Exam trap

The trap here is assuming that scaling up the capacity is always the solution, when the issue is actually the high frequency of operations causing throttling.

39
MCQmedium

You are a data engineer for a retail company using Microsoft Fabric. A data pipeline that ingests point-of-sale data from an on-premises SQL Server into a Fabric Lakehouse is intermittently failing. You need to identify the root cause of these failures and view detailed error messages for each pipeline run. Which feature should you use?

A.OneLake file explorer
B.Data Factory activity log in Azure Monitor
C.Capacity Metrics app
D.Monitor hub in Microsoft Fabric
AnswerD

The Monitor hub in Microsoft Fabric provides a centralized view of all activity across workspaces, including pipeline runs. It shows detailed execution history, status, and error messages for each run, allowing you to diagnose intermittent failures and identify the root cause without switching between multiple tools.

Why this answer

The Monitor hub is the central place in Microsoft Fabric to observe pipeline runs, including their status and detailed error messages. It aggregates activity across items and workspaces, making it the right tool to diagnose intermittent ingestion failures and see exactly what went wrong on each run.

Exam trap

The trap here is assuming that capacity-level monitoring tools such as the Capacity Metrics app provide per-run pipeline error details, when they actually focus on resource consumption and throttling.

40
Multi-Selectmedium

You are tasked with optimizing the performance of a Delta table in a Fabric Lakehouse that is queried frequently by both Spark notebooks and the SQL Analytics Endpoint. Which TWO techniques should you use to improve data skipping and read performance?

Select 2 answers
A.Enable V-Order on the Delta table.
B.Run the OPTIMIZE command on the table.
C.Convert the table to a CSV format.
D.Increase the Spark executor memory only.
E.Disable the use of shortcuts to the data.
AnswersA, B

V-Order is a write-time optimization that applies a special sorting and compression algorithm to Parquet files in OneLake. This makes them highly compatible with the Fabric compute engines, particularly the SQL Analytics Endpoint and Power BI, by enabling more efficient data skipping and faster decompression during query execution.

Why this answer

Option A (Enable V-Order on the Delta table) is correct because V-Order is a write-time optimization in Microsoft Fabric that reorganizes Parquet row groups with sorted, compressed data, enabling effective data skipping and faster reads across both Spark and the SQL Analytics Endpoint. Option B (Run the OPTIMIZE command on the table) is correct because OPTIMIZE compacts many small files into larger ones and, when combined with Z-ORDER or V-Order, improves file pruning and data skipping, directly boosting read performance. Option C is wrong because converting to CSV removes Delta's columnar Parquet storage, statistics, and transaction log, which destroys data skipping rather than improving it.

Option D is wrong because increasing Spark executor memory alone does not improve data skipping or I/O pruning and may not benefit the SQL Analytics Endpoint at all. Option E is wrong because disabling shortcuts would not enhance read performance and shortcuts are a valid, performant way to access data in OneLake.

Exam trap

DP-700 often tests the misconception that throwing more compute (executor memory, larger clusters) fixes slow Delta reads, when the real lever is reducing files scanned via OPTIMIZE, Z-ORDER, and V-Order.

41
MCQhard

You are troubleshooting a slow-running SQL query in a Fabric Data Warehouse. The query plan shows a high cost associated with a 'Broadcast Nested Loop Join'. What should you do to optimize this?

A.Increase the number of compute nodes.
B.Update the table statistics.
C.Disable the query optimizer.
D.Rewrite the query using a cursor.
AnswerB

Outdated statistics mislead the SQL engine into choosing sub-optimal join types. Updating them ensures the optimizer has an accurate view of table sizes and data distribution, leading to the selection of more appropriate join strategies like hash joins instead of nested loops.

Why this answer

A Broadcast Nested Loop Join is typically chosen when the optimizer lacks accurate statistics and underestimates row counts, leading it to broadcast a large table and loop. Updating table statistics gives the optimizer correct cardinality estimates so it can choose a more efficient join (e.g., hash or merge), which is the correct fix.

Exam trap

DP-700 often tests the instinct to 'add more compute' for any slow query, when the real fix is correcting optimizer inputs like statistics.

How to eliminate wrong answers

Option A is wrong because adding compute nodes does not fix a bad join algorithm chosen due to poor statistics; the plan remains inefficient. Option C is wrong because disabling the query optimizer removes the engine's ability to choose better plans, making performance worse. Option D is wrong because rewriting with a cursor forces row-by-row processing, which is far slower than set-based operations and is an anti-pattern in a warehouse.

42
MCQmedium

You are a data engineer managing a Microsoft Fabric workspace that contains a Data Warehouse. Users report that a specific Power BI report that queries the warehouse is slow, and you notice that the query frequently waits on 'SQL pools' resources. You need to identify which queries are consuming the most resources and causing contention. What should you do?

A.Run the 'Query Activity' report in the Fabric Data Warehouse and sort by CPU time or duration.
B.Use the Microsoft Fabric Capacity Metrics app to view the 'Timepoint' and 'Throttling' tabs.
C.Use the Fabric Monitoring hub to view the query activity and identify the top resource consumers.
D.Query the sys.dm_exec_requests and sys.dm_exec_sessions dynamic management views in the Data Warehouse.
AnswerA

The Query Activity report in Microsoft Fabric Data Warehouse provides detailed information about executed queries, including CPU time, duration, and resource usage. By sorting on CPU time or duration, you can identify the queries consuming the most resources and causing contention. This is the correct tool for monitoring and optimizing query performance in a Fabric Data Warehouse.

Why this answer

The Query Activity report in the Fabric Data Warehouse is specifically designed to provide insights into query execution, including resource consumption metrics such as CPU time and duration. By analyzing this report, you can pinpoint the queries that are consuming the most resources and causing contention, allowing you to optimize them. Other tools like the Capacity Metrics app or Monitoring hub provide capacity-level or job-level monitoring but lack the query-level detail required for this scenario.

Exam trap

The trap here is assuming that dynamic management views (DMVs) available in Azure Synapse Analytics are also available in Microsoft Fabric Data Warehouse, when in fact Fabric provides its own monitoring tools like the Query Activity report.

43
Multi-Selectmedium

You are optimizing a Microsoft Fabric Data Warehouse that supports a workload with a high volume of small, concurrent INSERT statements generated by an ingestion service. The Data Warehouse experiences slow write performance and frequent lock waits during peak hours. You need to improve write throughput and reduce contention. (Choose two.)

Select 2 answers
A.Disable the columnstore index on the target table and convert it to a heap.
B.Batch the small inserts into larger, less frequent load operations.
C.Enable read committed snapshot isolation on the Data Warehouse to remove all write blocking.
D.Increase the number of concurrent ingestion connections to parallelize the small inserts.
E.Use a staging table and load with CTAS or INSERT SELECT in set-based operations.
AnswersB, E

Each small INSERT creates its own transaction and writes new row groups, which fragments columnstore storage and increases overhead. Consolidating many small writes into fewer, larger batches reduces the number of transactions and row groups, lowers lock duration, and improves compression. This directly addresses both the slow write performance and the contention caused by the ingestion pattern.

Why this answer

The root cause is many tiny, concurrent write transactions that fragment columnstore storage and hold locks briefly but frequently. Batching inserts into larger loads and using a staging table with set-based CTAS or INSERT SELECT operations both reduce transaction count and produce better-organized row groups, improving throughput and lowering contention. Adding writers or removing columnstore storage does not fix the granularity problem.

Exam trap

The trap here is treating write contention as a parallelism problem and adding connections, when the actual fix is reducing the number and size granularity of transactions.

44
MCQhard

Refer to the exhibit. You are reviewing the capacity state for a production Fabric environment. Based on the JSON output from the monitoring API, what is the most likely impact on users and what should be your immediate action?

A.Interactive reports will stop working immediately; upgrade the SKU.
B.Background jobs are being delayed; optimize or reschedule ETL tasks.
C.The capacity is healthy; ignore the warning as smoothing is active.
D.Data loss is occurring in OneLake; check the transaction logs.
AnswerB

The exhibit shows significant Background Overage and active throttling. In Fabric, this means that background tasks have consumed more than their allotted share, and the system is now delaying new background operations to stay within the smoothed CU limits. Rescheduling these tasks to off-peak times is the best mitigation.

Why this answer

The JSON output indicates a capacity state where background operations are being throttled or delayed, which is characteristic of Fabric's capacity smoothing and throttling behavior when utilization exceeds the SKU's limits. Interactive reports are typically prioritized over background jobs, so they continue to function while ETL and other background tasks are queued or slowed. The immediate action is to optimize or reschedule ETL tasks to reduce load during peak periods, rather than upgrading the SKU or ignoring the warning.

Exam trap

DP-700 often tests the misconception that any capacity warning immediately impacts interactive reports, when in fact Fabric prioritizes interactive workloads and throttles background jobs first.

How to eliminate wrong answers

Option A is wrong because interactive reports are not the first to be affected; Fabric prioritizes interactive workloads over background jobs, so they would not stop working immediately. Option C is wrong because the warning indicates an active throttling state, not a healthy capacity; ignoring it would lead to further delays and potential failures. Option D is wrong because data loss does not occur in OneLake due to capacity throttling; OneLake provides durable storage, and throttling affects compute, not data integrity.

45
MCQmedium

You notice that your Delta table is experiencing slow read performance due to file fragmentation. Which command should you run to optimize the physical storage layout?

A.REORG TABLE
B.VACUUM TABLE
C.OPTIMIZE TABLE
D.ANALYZE TABLE
AnswerC

The OPTIMIZE command is specifically designed to compact small Delta files into larger files. This process significantly improves read performance by reducing the metadata volume and enabling more efficient disk I/O, which is essential for maintaining query speed in tables that undergo frequent updates or high-frequency ingestion.

Why this answer

OPTIMIZE TABLE is correct because it compacts many small Parquet files into fewer, larger files (target size ~1GB by default in Delta Lake), directly addressing file fragmentation and improving read performance. It also optionally performs Z-ORDER clustering to co-locate related data for data-skipping. This is the standard Delta Lake maintenance command for the small-files problem.

Exam trap

The trap here is confusing OPTIMIZE (compaction for read performance) with VACUUM (cleanup for storage reclamation) — candidates often pick VACUUM thinking it 'optimizes' the table, but it only deletes old files.

How to eliminate wrong answers

Option A is wrong because REORG TABLE is not a Delta Lake command — it is a Db2/DB2 LUW or mainframe utility for reorganizing table storage, not applicable to Delta tables in Fabric. Option B is wrong because VACUUM TABLE removes old, unreferenced files past the retention threshold (default 7 days) to reclaim storage; it does not merge small files and can actually reduce time-travel capability. Option D is wrong because ANALYZE TABLE collects statistics for the query optimizer (in some engines) but does not change the physical file layout or compact small files.

46
Multi-Selectmedium

You are optimizing a complex data pipeline in Microsoft Fabric that includes several Data Factory activities and Spark notebooks. You need to reduce the overall execution time and resource consumption. Which TWO actions should you take? (Choose two.)

Select 2 answers
A.Enable the 'Fast Copy' option in the Copy activity.
B.Set the notebook's retry policy to a high value.
C.Convert all source CSV files to Parquet with V-Order enabled.
D.Increase the timeout duration for all pipeline activities.
E.Use the 'Wait' activity between every notebook execution.
AnswersA, C

Fast Copy bypasses the traditional staging and transformation layers in some scenarios, allowing for direct and highly parallelized data movement between supported sources and sinks. This significantly reduces the time spent in the Copy activity and lowers the Compute Unit consumption required for the movement.

Why this answer

Option A is correct because enabling 'Fast Copy' in a Copy activity lets Microsoft Fabric bypass the traditional staging-based copy path and use optimized, high-throughput data movement, which directly reduces copy activity execution time. Option C is correct because converting CSV sources to Parquet with V-Order enabled provides columnar compression and Fabric's optimized file layout, improving read performance and reducing Spark notebook processing time and resource consumption. Option B is not appropriate because a high retry policy only helps recover from transient failures and does not reduce normal execution time or resource usage.

Option D is wrong because increasing activity timeouts merely allows longer-running activities and does not optimize performance. Option E is wrong because inserting Wait activities adds artificial delays and increases overall pipeline duration.

Exam trap

The trap is picking options that sound like 'safety nets' (retries, timeouts, waits) — these improve resilience or sequencing, not performance, and the exam rewards distinguishing optimization from mitigation.

47
Multi-Selecthard

You are monitoring a Microsoft Fabric Data Pipeline that ingests data from an on-premises SQL Server into a Lakehouse. The pipeline runs daily and occasionally fails due to transient network issues. You want to implement a robust error handling strategy to minimize manual intervention. Which two actions should you take? (Choose two.)

Select 2 answers
A.Add an If Condition activity after the Copy activity to check the status and send an alert on failure.
B.Enable logging and set up alerts in the Monitor hub for pipeline failures.
C.Configure a retry policy on the Copy activity with a suitable retry count and interval.
D.Set the pipeline's concurrency to 1 to avoid conflicts.
E.Use a ForEach activity to iterate over each row and retry individually.
AnswersB, C

Enabling logging and alerts in the Monitor hub ensures that failures are captured and notifications are sent. While this does not automatically retry, it is essential for monitoring and quick response. Combined with a retry policy, it forms a robust error handling strategy that minimizes manual intervention by alerting only when retries are exhausted.

Why this answer

Configuring a retry policy on the Copy activity automatically handles transient failures, while enabling logging and alerts ensures visibility when retries fail. Together, they reduce manual intervention by automating recovery and notifying only when necessary. The other actions do not provide automatic retry or are inefficient.

Exam trap

The trap here is assuming that adding conditional checks or alerts alone will recover from failures, when automatic retry is needed for transient errors.

48
Multi-Selectmedium

When monitoring Fabric Data Pipelines, which TWO metrics are most useful for identifying performance issues in data copy activities?

Select 2 answers
A.Data throughput (MB/s).
B.Number of active users in the workspace.
C.Activity duration.
D.Total number of rows in the destination table.
E.The name of the user who triggered the pipeline.
AnswersA, C

Throughput is the primary indicator of how efficiently data is being transferred between the source and destination. Low throughput often points to bottlenecks such as network bandwidth, source system limitations, or inefficient data format conversion during the copy process.

Why this answer

Data throughput (MB/s) (option A) is correct because it directly measures how fast data is being moved during a copy activity, so a drop in throughput signals a bottleneck in the source, sink, network, or integration runtime. Activity duration (option C) is correct because it shows how long the copy activity takes to complete, and comparing it against expected baselines reveals slowdowns or degradation in copy performance. Together, these two metrics isolate whether a copy activity is slow because of low transfer rate or excessive elapsed time.

The other options do not belong: the number of active workspace users (B) and the triggering user's name (E) are usage/audit details unrelated to copy performance, and the total row count in the destination table (D) reflects data volume, not the speed or efficiency of the copy operation.

Exam trap

DP-700 often tests the confusion between volume metrics (row counts, table size) and performance metrics (throughput, duration), so candidates must select metrics that measure speed and efficiency, not data quantity.

49
MCQeasy

You are monitoring a Dataflow Gen2 refresh in Microsoft Fabric. The refresh fails with an error: 'Mashup Evaluation Error'. What is the most common cause of this error when working with large datasets in Dataflows?

A.The destination Lakehouse has reached its storage limit.
B.The transformation logic is too memory-intensive for the evaluator.
C.The Spark pool assigned to the Dataflow is offline.
D.The user does not have 'Execute' permissions on the Dataflow.
AnswerB

Complex steps like 'Group By' on high-cardinality columns, multiple joins, or large 'Sort' operations can exceed the memory allocated to the Power Query mashup container. Simplifying the steps or using 'staged queries' to break up the logic can often resolve these types of evaluation errors.

Why this answer

A 'Mashup Evaluation Error' in Dataflow Gen2 indicates the Power Query mashup engine failed while evaluating the transformation graph — most commonly because the transformation logic (joins, group-bys, custom columns) is too memory-intensive for the evaluator to process on large datasets. The mashup engine runs in a constrained container, so memory pressure during evaluation is the typical root cause.

Exam trap

DP-700 often tests whether candidates understand that Dataflow Gen2 uses the Power Query mashup engine (not Spark), tempting them to blame Spark pool availability or storage limits for what is actually an in-memory evaluation failure.

How to eliminate wrong answers

Option A is wrong because a Lakehouse storage limit would produce a write/destination error, not a mashup evaluation error during transformation. Option C is wrong because Dataflow Gen2 does not use a Spark pool — it uses the Power Query mashup engine, so a Spark pool being offline is not applicable. Option D is wrong because a permissions failure would surface as an authorization/access-denied error, not a mashup evaluation error.

50
MCQeasy

A data engineer in your team frequently runs interactive notebooks against a Fabric Lakehouse and reports that queries against a large Delta table are slow, even though the data volume has not changed. You suspect that the table's small files are degrading read performance. Which action should you take to address this?

A.Run VACUUM on the Delta table to remove old file versions and reclaim storage.
B.Run the OPTIMIZE command on the Delta table to compact small files into larger ones.
C.Increase the number of partitions in the table by adding a new partition column.
D.Enable Z-ORDER on the table by clustering on a frequently filtered column.
AnswerB

OPTIMIZE compacts many small Parquet files into fewer larger files, which reduces per-file overhead and improves scan performance for read-heavy workloads. Since the symptom is slow reads on a large Delta table whose volume is stable, file compaction directly targets the likely cause, making this the appropriate corrective action.

Why this answer

Compacting small files with OPTIMIZE reduces the number of files scanned and the associated per-file overhead, which is the standard remedy when a Delta table has accumulated many small files and reads have slowed. Vacuuming addresses storage reclamation, and partitioning or Z-ORDER address pruning, none of which resolve the small-file scan overhead directly.

Exam trap

The trap here is confusing storage-reclamation maintenance such as VACUUM with read-performance maintenance such as file compaction.

51
MCQeasy

You want to automate the monitoring of your Fabric Capacity and receive a notification on your mobile device whenever the capacity utilization exceeds 90% for more than 15 minutes. Which Microsoft Fabric feature should you use to implement this alert?

A.Data Activator
B.Azure Service Health
C.Power BI Subscription
D.OneLake File Explorer
AnswerA

Data Activator allows you to create 'Reflex' items that monitor data streams or capacity metrics. You can define a rule for 90% utilization and set a condition for the duration. When the condition is met, it can trigger actions such as sending notifications or starting a pipeline.

Why this answer

Data Activator is the Fabric feature designed to monitor real-time data streams and trigger alerts or actions when conditions are met. It can watch a capacity metric, detect when utilization exceeds 90% for 15 minutes, and send a mobile notification via Teams, email, or Power Automate. It is purpose-built for this event-driven alerting scenario.

Exam trap

DP-700 often tests the confusion between scheduled report delivery (Power BI Subscriptions) and condition-based alerting (Data Activator), tempting candidates to pick the familiar Power BI option.

How to eliminate wrong answers

Option B is wrong because Azure Service Health reports on Azure platform incidents and maintenance, not on Fabric capacity utilization thresholds. Option C is wrong because a Power BI Subscription emails a report snapshot on a schedule; it cannot evaluate a 90%-for-15-minutes condition or send mobile push alerts. Option D is wrong because OneLake File Explorer is a desktop tool for browsing OneLake files, with no monitoring or alerting capability.

52
MCQmedium

You are a data engineer at a retail company using Microsoft Fabric. Your team runs a nightly data pipeline that loads sales data into a Lakehouse and then refreshes a Power BI semantic model. The pipeline has been missing its completion SLA. You use the Monitoring hub to inspect recent runs and notice that the 'Dataflow Gen2' activity consistently takes over two hours, while other activities finish quickly. You need to reduce the overall pipeline duration. What should you do first?

A.Optimize the Dataflow Gen2 by reducing unnecessary transformations and enabling fast copy.
B.Convert the Dataflow Gen2 activity to a Spark notebook and run it on a larger cluster.
C.Increase the capacity size of the Fabric capacity assigned to the workspace.
D.Enable query caching on the Lakehouse used by the Dataflow Gen2 activity.
AnswerA

Dataflow Gen2 performance is often impacted by complex transformations and lack of fast copy. Reducing unnecessary steps and enabling fast copy allows the dataflow to leverage optimized data movement and query folding, significantly cutting runtime. This directly addresses the slow activity identified in the Monitoring hub, making it the most effective first step.

Why this answer

The Dataflow Gen2 activity is the clear bottleneck. Optimizing its internal logic by removing unnecessary transformations and enabling fast copy leverages built-in performance features and directly reduces runtime. Other options either do not target the specific activity or involve larger architectural changes that are not the immediate first step.

Exam trap

The trap here is assuming that scaling up capacity will automatically fix a single slow activity, when the real issue is inefficient dataflow logic.

53
MCQmedium

A Power BI report using a semantic model in 'Direct Lake' mode is performing poorly. You want to determine if the queries are falling back to 'DirectQuery' mode due to memory constraints or unsupported DAX features. Which tool is best suited for this specific optimization task?

A.DAX Studio
B.Spark UI
C.SQL Server Management Studio (SSMS) Activity Monitor
D.OneLake Explorer
AnswerA

DAX Studio can connect to the semantic model's XMLA endpoint and capture trace events. It allows you to see if a query was resolved using the Direct Lake fast path or if it fell back to DirectQuery mode, which is significantly slower and puts load on the SQL Analytics Endpoint.

Why this answer

DAX Studio connects directly to a Power BI semantic model and exposes query traces, including whether a Direct Lake query fell back to DirectQuery. It shows the fallback reason (memory limit, unsupported DAX, or SQL view) and the actual query plan, making it the right tool for this optimization.

Exam trap

The trap here is confusing monitoring tools across Fabric workloads — candidates may pick Spark UI or SSMS because they are familiar monitoring interfaces, but only DAX Studio exposes Direct Lake fallback diagnostics.

How to eliminate wrong answers

Option B is wrong because Spark UI monitors Spark jobs in Fabric notebooks and lakehouses, not Power BI semantic model query behavior. Option C is wrong because SSMS Activity Monitor targets SQL Server engine activity, not Power BI Direct Lake fallback diagnostics. Option D is wrong because OneLake Explorer is a file browsing tool for OneLake data, not a semantic model query analyzer.

54
Multi-Selectmedium

You are a data engineer at a logistics company using Microsoft Fabric. You are monitoring a Fabric pipeline that copies data from an Azure SQL Database to a Lakehouse. The pipeline occasionally runs longer than expected. You need to identify the causes of the increased duration. Which two actions should you perform? (Choose two.)

Select 2 answers
A.Reconfigure the pipeline to use a staging area in Azure Blob Storage before loading into the Lakehouse.
B.Enable diagnostic settings to stream pipeline logs to Azure Log Analytics and analyze query performance.
C.Use the Capacity Metrics app to check for throttling or overload on the Fabric capacity.
D.Review the pipeline's activity run details in the Monitoring hub to identify the slowest activity.
E.Check the Data Integration Units (DIU) usage and adjust the copy activity parallelism.
AnswersC, D

The Microsoft Fabric Capacity Metrics app shows resource consumption and throttling events. If the capacity is overloaded, pipelines may run slower due to resource contention. Checking this app helps determine whether the increased duration is due to capacity limits, which is a common cause of performance degradation in shared environments.

Why this answer

To diagnose increased pipeline duration, you should first review the activity run details in the Monitoring hub to identify the slowest activity, and then check the Capacity Metrics app to see if capacity throttling is impacting performance. These two actions provide direct insight into both the pipeline's internal bottleneck and the external resource constraints.

Exam trap

The trap here is jumping to optimization techniques like adjusting DIU or adding staging areas without first identifying the actual cause through monitoring and capacity metrics.

55
MCQhard

In Microsoft Fabric, you notice that a background job (such as a Dataflow refresh) is taking significantly longer than usual, even though the total Capacity Unit (CU) usage is below the limit. What is the most likely cause of this behavior based on Fabric's capacity management rules?

A.The background job has been rejected by the system.
B.The job is being smoothed over a 24-hour window.
C.The SQL Analytics Endpoint has higher priority than Dataflows.
D.The workspace is in a 'Trial' capacity which lacks background processing.
AnswerB

Fabric's smoothing policy means that the impact of a heavy background job is spread out over 24 hours. If there was a large spike in usage earlier, the system may delay the allocation of CUs to new background tasks to stay within the average limit, causing them to take longer to complete.

Why this answer

Fabric smooths background jobs over a 24-hour window, meaning a job's CU consumption is averaged across the day rather than executed at full speed immediately. When the smoothed rate is throttled, the job runs slower even though total CU usage appears below the capacity limit. This is why a Dataflow refresh can take much longer than usual despite seemingly available capacity.

Exam trap

DP-700 often tests the misconception that low total CU usage means no throttling — candidates forget that background jobs are smoothed over 24 hours and can be throttled even when the capacity appears underutilized.

How to eliminate wrong answers

Option A is wrong because rejected jobs fail outright rather than run slowly — rejection is a hard stop, not a slowdown. Option C is wrong because Fabric does not give the SQL Analytics Endpoint priority over Dataflows; interactive vs. background is the relevant distinction, not endpoint type. Option D is wrong because Trial capacities do support background processing; they simply have limited CU and specific throttling behavior.

56
MCQhard

You are a data engineer using Microsoft Fabric. A Spark notebook reads a large Delta table, performs a groupBy aggregation, and writes the result to a Lakehouse. You notice that the job is spilling to disk heavily and taking a long time. The Spark UI shows that the stage with the aggregation has many tasks and high shuffle read/write. You need to reduce shuffle and improve performance. What should you do?

A.Cache the Delta table in memory before performing the aggregation.
B.Enable Adaptive Query Execution (AQE) and set spark.sql.adaptive.coalescePartitions.enabled to true.
C.Repartition the Delta table by the groupBy key before the aggregation.
D.Increase the number of shuffle partitions by setting spark.sql.shuffle.partitions to a higher value.
AnswerB

Adaptive Query Execution (AQE) dynamically optimizes shuffle partitions by coalescing small partitions and handling skew. Enabling AQE with partition coalescing reduces the number of tasks and mitigates spilling. In this scenario, AQE can automatically adjust the shuffle partition size based on runtime statistics, improving performance without manual tuning.

Why this answer

Adaptive Query Execution (AQE) is designed to optimize shuffle operations at runtime. By enabling AQE and partition coalescing, Spark can dynamically merge small partitions and handle skew, reducing spilling and improving aggregation performance. Other options either do not address the root cause or may introduce additional overhead.

Exam trap

The trap here is thinking that increasing shuffle partitions always helps, when the problem may be too many small partitions or skew.

57
MCQhard

A data engineer maintains a Fabric Data Pipeline that loads a Lakehouse table every 15 minutes. The pipeline recently began exceeding its expected duration, and downstream reports are stale. You review the pipeline run history and see that the Copy activity reads a source folder containing both new files and files that were already processed in prior runs. You need to ensure each run processes only newly arrived files and completes faster without changing the source system. What should you configure?

A.Enable the fault tolerance setting to skip incompatible rows and log them.
B.Use the Copy activity with a modified datetime filter on the source folder to select only files newer than the last run.
C.Set the Copy activity parallelism to a higher degree and enable staging.
D.Configure a tumbling window trigger with the schedule aligned to file timestamps.
AnswerB

Applying a datetime filter to the source folder restricts the file enumeration to items whose modified timestamp is newer than the last successful run. The Copy activity then reads only newly arrived files, eliminating redundant work and shortening the run. This approach changes neither the source system nor the file layout, matching the requirement precisely.

Why this answer

The pipeline is slow because each run enumerates and copies files that were already processed. Filtering the source folder by modified datetime so only files newer than the last run are selected removes the redundant reads and shortens execution. Parallelism, fault tolerance, and trigger scheduling do not restrict which files are read, so they cannot solve the underlying issue.

Exam trap

The trap here is assuming that a trigger schedule or higher parallelism automatically limits processing to new files, when the file selection logic in the Copy activity is what actually controls scope.

58
MCQmedium

You are monitoring a Microsoft Fabric Lakehouse. You observe that queries against a specific Delta table are performing poorly. You identify that the table contains a large number of small files. Which command should you run to optimize the file layout?

A.ALTER TABLE REFRESH
B.VACUUM
C.OPTIMIZE table_name
D.ANALYZE TABLE
AnswerC

Running OPTIMIZE reorganizes the data by compacting small files into larger, optimally sized Parquet files. This process improves read performance by reducing the number of files the engine must scan, which is the standard procedure for mitigating small file issues in Lakehouses.

Why this answer

OPTIMIZE table_name is the correct Delta Lake command to compact many small files into larger, more efficient files, directly addressing the small-file problem that degrades query performance. It performs bin-packing to coalesce files toward a target size (default 1 GB in Databricks/Fabric), reducing file-open overhead and improving scan throughput. Optionally, OPTIMIZE ...

ZORDER BY can co-locate related data for further gains.

Exam trap

DP-700 often tests the distinction between Delta maintenance commands — candidates confuse VACUUM (cleanup of old files) with OPTIMIZE (compaction of small files), or pick ANALYZE TABLE thinking statistics collection improves layout, when only OPTIMIZE physically reorganizes data.

How to eliminate wrong answers

Option A is wrong because ALTER TABLE REFRESH is not a valid Delta Lake optimization command — it does not exist for this purpose and would not compact files. Option B is wrong because VACUUM removes old, unreferenced data files after the retention period (default 7 days) to reclaim storage; it does not merge small files and can even delete files still needed for time travel. Option D is wrong because ANALYZE TABLE collects statistics for the query optimizer (used by cost-based optimization) but does not physically reorganize or compact the underlying files.

59
Multi-Selectmedium

You are responsible for monitoring a Microsoft Fabric analytics solution that includes Data Pipelines, a Data Warehouse, and a Lakehouse. You need to detect data quality regressions and operational failures with minimal manual effort. Which two actions should you take? (Choose two.)

Select 2 answers
A.Increase the Fabric capacity SKU so that monitoring queries and pipeline activities run with more available compute.
B.Instruct each analyst to review the Lakehouse table preview each morning and report anything that looks incorrect.
C.Create a Data Warehouse query that counts null and out-of-range values, then schedule it and trigger an alert when thresholds are exceeded.
D.Disable automatic retries on all pipeline activities so that failures surface immediately in the run history.
E.Enable and configure Data Activator alerts on Data Pipeline activity failures so the responsible team is notified automatically.
AnswersC, E

A scheduled validation query against the warehouse produces a repeatable, objective measure of data quality, such as null rates or values outside expected ranges. Pairing it with an alert converts that measurement into proactive notification when a regression occurs. This is the low-effort, automated way to catch data quality drift without embedding checks into every individual load.

Why this answer

Automated detection requires both an alerting mechanism for operational failures and a repeatable measurement of data quality. Data Activator alerts on pipeline activity failures cover the orchestration side, while a scheduled validation query with threshold-based alerting covers the data side. Together they provide continuous, low-effort coverage across the pipeline, warehouse, and lakehouse components.

Exam trap

The trap here is equating a larger capacity or disabled retries with better monitoring, when neither actually notifies anyone or measures data correctness.

60
MCQhard

You are a data engineer optimizing a Fabric Data Warehouse. A critical report queries a large fact table with filters on a date column and a region column. You notice that the query frequently performs a full table scan. You decide to create a clustered columnstore index on the fact table. Which statement about clustered columnstore indexes in Fabric Data Warehouse is correct?

A.Clustered columnstore indexes automatically create statistics on all columns, eliminating the need for manual statistics maintenance.
B.Clustered columnstore indexes must be rebuilt after every data load to remain effective.
C.Clustered columnstore indexes in Fabric Data Warehouse cannot be created on tables with fewer than one million rows.
D.Clustered columnstore indexes store data in a columnar format and can improve compression and query performance for analytical workloads.
AnswerD

Clustered columnstore indexes store data column-wise, which provides high compression and faster scans for analytical queries that aggregate or filter on a subset of columns. This makes them ideal for large fact tables in Fabric Data Warehouse, reducing I/O and improving performance.

Why this answer

Clustered columnstore indexes store data in a columnar format, offering excellent compression and fast analytical query performance by scanning only relevant columns. They are a core optimization for large fact tables in Fabric Data Warehouse, reducing I/O for aggregation-heavy reports.

Exam trap

The trap here is assuming that columnstore indexes eliminate the need for statistics maintenance or require constant rebuilds, when in reality they still rely on statistics and use delta rowgroups for incremental loads.

61
MCQhard

A Microsoft Fabric Data Pipeline loads files into a Lakehouse every 15 minutes. You observe that runs sometimes overlap, and a downstream Dataflow Gen2 refresh occasionally reads a partially written table. You need to prevent overlapping execution and guarantee that the Dataflow Gen2 activity starts only after the entire copy activity has committed its output. What should you configure?

A.Configure the copy activity to write to a staging folder and enable the Dataflow Gen2 activity's automatic retry on failure.
B.Increase the copy activity's degree of parallelism and schedule the Dataflow Gen2 activity on a separate trigger offset by five minutes.
C.Set the pipeline concurrency to 1 and add a dependency from the Dataflow Gen2 activity to the copy activity on the Succeeded condition.
D.Add a Wait activity before the Dataflow Gen2 activity that pauses for the expected copy duration, then continue the pipeline.
AnswerC

Limiting pipeline concurrency to one prevents a new scheduled run from starting while the previous run is still executing, which eliminates the overlapping writes. Chaining the Dataflow Gen2 activity to the copy activity with a Succeeded dependency ensures the refresh begins only after the copy activity has fully committed, so no partial table is read.

Why this answer

The overlap comes from multiple scheduled runs coexisting, which is solved by constraining pipeline concurrency to a single instance. The partial-read risk comes from ordering, which is solved by a Succeeded dependency so the Dataflow Gen2 activity cannot begin until the copy activity has completed and committed its output. Together they enforce both mutual exclusion and correct sequencing.

Exam trap

The trap here is treating a fixed delay or a retry policy as equivalent to a completion dependency, when only a success condition actually guarantees the upstream write has committed.

62
MCQeasy

Which Fabric tool allows you to visually track the total compute usage of your workspace over a specific timeframe?

A.OneLake monitoring dashboard
B.Capacity Metrics app
C.Pipeline monitoring view
D.Azure Advisor
AnswerB

The Capacity Metrics app is the dedicated tool in Fabric for tracking compute utilization. It offers detailed views into how much capacity your workloads are consuming, allowing you to monitor usage trends, identify peak periods, and ensure that your capacity is appropriately sized for your organizational demand.

Why this answer

Capacity Metrics app is correct because it is the dedicated Fabric tool that provides detailed visibility into compute usage (CU consumption) across workspaces, items, and time periods. It helps capacity admins track utilization, throttling, and overage, enabling informed scaling decisions.

Exam trap

The trap is confusing monitoring tools: candidates may pick Pipeline monitoring (for pipeline runs) or Azure Advisor (for Azure recommendations) instead of the Capacity Metrics app, which is specifically for compute usage tracking.

How to eliminate wrong answers

Option A is wrong because there is no 'OneLake monitoring dashboard' as a distinct Fabric tool for compute usage — OneLake is the storage layer, and monitoring is done via the Capacity Metrics app. Option C is wrong because Pipeline monitoring view shows individual pipeline run details, not aggregated workspace compute usage over time. Option D is wrong because Azure Advisor provides recommendations for Azure resources, not Fabric capacity compute usage tracking.

63
MCQhard

You are monitoring a Microsoft Fabric Data Pipeline that copies data from an on-premises SQL Server to a Lakehouse. The pipeline occasionally fails with a timeout error during the copy activity. You need to identify the root cause and reduce the chance of future timeouts. What should you do first?

A.Review the copy activity's error details and the on-premises data gateway logs.
B.Increase the DIU (Data Integration Unit) setting for the copy activity.
C.Enable staging for the copy activity to use a temporary Azure Blob Storage location.
D.Reduce the copy activity's parallelism by setting the degree of copy parallelism to 1.
AnswerA

The copy activity error details and gateway logs provide the specific reason for the timeout, such as network instability, gateway resource exhaustion, or query performance on the source. Examining these logs first ensures you address the actual root cause rather than applying a generic fix. This is the correct diagnostic step before making configuration changes.

Why this answer

The copy activity error details and on-premises data gateway logs contain the specific reason for the timeout, such as network latency, gateway resource limits, or source query performance. Reviewing these first allows you to target the actual cause. Changing DIUs, enabling staging, or reducing parallelism are optimization steps that should follow a proper diagnosis.

Exam trap

The trap here is jumping to performance tuning options like increasing DIUs or enabling staging without first reading the error details and gateway logs to understand why the timeout occurred.

64
MCQeasy

A Fabric workspace uses a pipeline that copies data from Azure SQL Database into a Lakehouse. The pipeline succeeds, but row counts in the Lakehouse are consistently lower than in the source. You need to identify which rows were skipped during the copy. Which pipeline feature should you inspect?

A.The Lakehouse table's Delta transaction log, which records each write and its row count
B.The Azure SQL Database query store, which captures execution plans for source queries
C.The pipeline's schedule trigger history, which shows when runs started and ended
D.The pipeline run's activity output, which includes rows read, rows written, and rows skipped with error details
AnswerD

Copy activity output reports rows read, rows written, and rows skipped, along with error rows and reasons when fault tolerance is enabled. This directly shows how many rows were dropped and why, such as data type conversion failures. Inspecting it is the fastest way to identify skipped rows without changing the pipeline.

Why this answer

When a copy activity completes but writes fewer rows than it reads, the difference is explained by skipped rows, typically from fault tolerance settings or type conversion errors. The activity output lists rows read, written, and skipped with error detail, making it the authoritative place to see exactly which records were excluded and why. Other monitoring surfaces lack this data-level view.

Exam trap

The trap here is trusting a green pipeline status as proof that all rows were copied, when skipped rows can occur inside a successful run.

65
MCQhard

Your organization uses a Fabric Spark notebook to process large volumes of streaming data into a Delta table. You notice that over time, query performance on the table is degrading significantly. Upon investigation, you find thousands of small files in the underlying OneLake folder. Which optimization strategy should you implement to resolve the performance issue while maintaining data integrity?

A.Run the VACUUM command with a retention period of zero hours.
B.Increase the Spark executor count to distribute the load.
C.Execute the OPTIMIZE command on the Delta table.
D.Disable V-Order on the Spark session to reduce write overhead.
AnswerC

The OPTIMIZE command performs data compaction by merging small files into larger, more optimal Parquet files while preserving the transaction log's integrity. This process reduces the number of file metadata operations required during a read, which directly addresses the root cause of the observed query performance degradation.

Why this answer

The OPTIMIZE command on a Delta table compacts thousands of small Parquet files into larger, more efficient files, dramatically improving read performance without altering the underlying data. It preserves ACID guarantees and can optionally apply Z-ORDER for multi-dimensional clustering. This directly addresses the small-files problem caused by frequent streaming micro-batches.

Exam trap

DP-700 often tests the confusion between OPTIMIZE (compaction for read performance) and VACUUM (garbage collection for storage reclamation), so candidates must recognize that small-file performance issues require compaction, not deletion.

How to eliminate wrong answers

Option A is wrong because VACUUM with a retention period of zero hours deletes all old file versions, breaking time travel and potentially corrupting concurrent readers — it removes files, not compacts them, so it does not solve the small-files problem. Option B is wrong because increasing executor count adds compute parallelism but does not reduce the number of small files; the bottleneck is file metadata overhead, not CPU. Option D is wrong because disabling V-Order removes an optimization that improves Parquet read performance in Fabric; it would degrade, not improve, query performance.

66
MCQeasy

You are a data engineer working with a Microsoft Fabric Lakehouse. You need to monitor the health of your data ingestion processes and receive alerts when a pipeline fails. Which feature should you use to set up proactive notifications?

A.Power BI data alerts
B.Fabric capacity metrics alerts
C.OneLake diagnostics logs
D.Data Activator in Microsoft Fabric
AnswerD

Data Activator allows you to define triggers and alerts based on conditions in your data or pipeline events. You can configure it to send notifications when a pipeline fails, providing proactive monitoring without manual checks. It integrates with Teams, email, and other channels for timely alerts.

Why this answer

Data Activator is the Microsoft Fabric feature designed to monitor data and events, including pipeline runs, and trigger alerts or actions. By setting up a trigger on pipeline failure events, you can receive proactive notifications through Teams or email, ensuring rapid response to ingestion issues.

Exam trap

The trap here is confusing Power BI data alerts, which monitor dashboard visuals, with the broader event-driven alerting capabilities of Data Activator for pipeline failures.

67
MCQmedium

You need to minimize the performance impact of concurrent data refreshes on your Fabric capacity. What is the best strategy?

A.Run all pipelines simultaneously at 00:00
B.Increase the refresh frequency for all items
C.Stagger scheduled refreshes to distribute load
D.Limit access to the workspace for users
AnswerC

Staggering schedules flattens the load profile on your capacity. By ensuring that intensive data refreshes do not overlap, you maximize throughput and minimize the chance of hitting compute limits, ensuring that all jobs have enough resources to finish within their windows.

Why this answer

Staggering scheduled refreshes distributes the compute and memory load across time, preventing multiple heavy refresh operations from competing for the same Fabric capacity resources simultaneously. Fabric capacities have finite compute units (CUs), and concurrent refreshes can cause throttling, queuing, and degraded performance for interactive users. Spreading refreshes across the day or night smooths the load curve and keeps the capacity within its CU budget.

Exam trap

DP-700 often tests capacity management intuition — candidates pick 'run everything at midnight' thinking off-hours solves contention, but the exam expects recognition that concurrency, not clock time, drives capacity pressure, and staggering is the correct mitigation.

How to eliminate wrong answers

Option A is wrong because running all pipelines simultaneously at midnight concentrates load into a single spike, maximizing contention and likely triggering capacity throttling — the opposite of the goal. Option B is wrong because increasing refresh frequency for all items multiplies the total workload and concurrency, worsening the performance impact rather than minimizing it. Option D is wrong because limiting workspace access for users does not reduce the compute cost of data refreshes; it only restricts who can view results and is unrelated to refresh load management.

68
MCQeasy

You are monitoring long-running Spark jobs in your Fabric workspace. You want to identify which specific stages of the job are consuming the most time. Which tool should you use?

A.Fabric Capacity Metrics app
B.Spark UI
C.OneLake file explorer
D.Azure Monitor logs
AnswerB

The Spark UI provides an in-depth view of job execution, including DAG visualizations, stage durations, and task-level metrics. It is the primary tool for investigating performance bottlenecks within Spark jobs, allowing developers to drill down into why specific operations are taking longer than expected during execution.

Why this answer

The Spark UI is the purpose-built diagnostic interface that exposes job, stage, and task-level timing, including the DAG visualization, stage durations, shuffle read/write metrics, and skew indicators. When you need to pinpoint which specific stages of a long-running Spark job are consuming the most time, the Spark UI's Stages tab is the canonical tool.

Exam trap

The trap is assuming the Fabric Capacity Metrics app is the go-to for all Fabric performance tuning — it is excellent for capacity/CU consumption but useless for stage-level Spark job diagnosis, which only the Spark UI provides.

How to eliminate wrong answers

Option A is wrong because the Fabric Capacity Metrics app reports workspace and capacity-level consumption (CU usage, throttling, item-level utilization) — it does not break down Spark job execution into individual stages or tasks. Option C is wrong because OneLake file explorer is a Windows client for browsing and managing OneLake files, with no job telemetry capability. Option D is wrong because Azure Monitor logs capture platform-level diagnostics and metrics but do not provide the fine-grained Spark stage/task DAG and shuffle-level timing that the Spark UI exposes.

69
MCQhard

Refer to the exhibit. You are reviewing the execution plan for a query running on a Fabric SQL Analytics Endpoint. The query is performing slower than expected. Based on the JSON snippet of the plan, what is the most likely cause of the performance bottleneck?

A.The RemoteScan is failing to find the Parquet files in OneLake.
B.The BroadcastExchange is moving too much data for a large table join.
C.The HashJoin is using an unsupported data type for the join condition.
D.OneLake storage is currently in a read-only state due to maintenance.
AnswerB

BroadcastExchange is efficient for small tables but becomes a bottleneck if the table is large, as it sends a full copy of the data to every compute node. The high output row count indicates that a significant volume of data is being processed, which can saturate the network and degrade performance.

Why this answer

In Fabric SQL Analytics Endpoint execution plans, a BroadcastExchange operation is used to distribute a small table to all nodes for a join. If the table being broadcast is actually large, it causes excessive data movement and memory pressure, leading to performance bottlenecks. The JSON snippet likely shows a BroadcastExchange with high data volume, indicating a misestimation by the optimizer.

Exam trap

DP-700 often tests the ability to interpret execution plan operators; candidates may focus on storage or syntax errors when the real issue is data movement due to a poor join strategy.

How to eliminate wrong answers

Option A is wrong because a RemoteScan failure would result in an error, not a performance bottleneck; the query would fail entirely. Option C is wrong because HashJoin supports common data types; an unsupported type would cause a compilation error, not slowness. Option D is wrong because a read-only OneLake state would cause write failures, not slow query performance.

70
MCQeasy

A Microsoft Fabric workspace hosts several notebooks that write to a Lakehouse. A data engineer notices that one notebook occasionally fails with an error stating that a Delta table's transaction log conflicts with a concurrent write from a different notebook. You need to review the history of operations performed on that Delta table to identify which writers overlapped. Which capability should you use?

A.The Spark application UI for the failed notebook, focusing on the failed stage.
B.The Monitoring hub, filtered to notebook items in the workspace.
C.The Lakehouse explorer, where you inspect the table's column schema and file listing.
D.The Delta table history, queried with DESCRIBE HISTORY against the table.
AnswerD

DESCRIBE HISTORY returns the Delta transaction log entries for a table, including each operation, the user or job that performed it, and timestamps. Reviewing those entries shows which writes overlapped and produced the concurrency conflict. This is the authoritative record of table-level operations and is exactly suited to identifying competing writers.

Why this answer

Delta table history, retrieved with DESCRIBE HISTORY, enumerates each committed operation against the table along with timestamps and the writer responsible. That chronology makes overlapping writes from separate notebooks visible, which is precisely what is needed to explain the transaction log conflict and to decide how to serialize the writers.

Exam trap

The trap here is using notebook-level run monitoring to explain a table-level concurrency conflict, when the Delta transaction log is the record that names the competing write operations.

71
MCQeasy

You are monitoring a Microsoft Fabric Data Pipeline that runs daily. The pipeline has multiple activities, and you need to quickly identify which activity failed and view its error message. Which feature should you use?

A.The Monitoring hub, filtering by pipeline run ID.
B.The Capacity Metrics app, filtering by item type.
C.The workspace lineage view, selecting the pipeline.
D.The Data Factory activity log in Azure Monitor.
AnswerA

The Monitoring hub in Microsoft Fabric provides a centralized view of all activity runs, including pipelines. By filtering for the specific pipeline run ID, you can see the status of each activity, including failures, and drill into error details. This is the primary tool for monitoring and troubleshooting pipeline executions in Fabric.

Why this answer

The Monitoring hub is the central place in Microsoft Fabric to view pipeline runs, activity statuses, and error messages. Filtering by pipeline run ID allows quick identification of the failed activity and its error details. Other options do not provide per-activity execution diagnostics within Fabric.

Exam trap

The trap here is assuming that the Capacity Metrics app provides pipeline activity-level error details, when it only shows capacity usage.

72
MCQmedium

A Microsoft Fabric tenant is experiencing frequent throttling of background operations during the early morning hours. You have been tasked with identifying which specific items are consuming the most capacity units (CU) to determine if a workload needs to be rescheduled. Which tool should you use to get the most granular view of item-level CU consumption over a 14-day period?

A.The Microsoft 365 Admin Center usage reports.
B.The Microsoft Fabric Capacity Metrics app.
C.Azure Monitor Log Analytics workspace queries.
D.The Workspace settings 'Usage Metrics' tab.
AnswerB

This application provides detailed visual insights into how individual items like notebooks, pipelines, and semantic models consume capacity over time. It allows data engineers to drill down into specific time intervals to see the impact of background versus interactive operations and identify specific items responsible for capacity exhaustion.

Why this answer

The Microsoft Fabric Capacity Metrics app is purpose-built for capacity monitoring and provides item-level CU consumption breakdowns over a rolling 14-day window, which is exactly what's needed to identify which specific items are driving throttling. It surfaces interactive vs. background usage, throttling events, and per-item consumption, making it the correct tool for granular, item-level analysis.

Exam trap

DP-700 often tests the distinction between tenant-wide capacity telemetry (Capacity Metrics app) and workspace-scoped usage reporting (Usage Metrics tab), tempting candidates to pick the workspace-level tool for a tenant-level throttling investigation.

How to eliminate wrong answers

Option A is wrong because Microsoft 365 Admin Center usage reports cover M365 service activity (Exchange, Teams, SharePoint), not Fabric capacity CU consumption. Option C is wrong because Azure Monitor Log Analytics can ingest Fabric diagnostic logs, but it does not natively render item-level CU consumption views over 14 days without custom KQL authoring and does not provide the built-in capacity metrics semantics. Option D is wrong because the Workspace 'Usage Metrics' tab reports view/usage statistics for Power BI artifacts in a single workspace, not capacity-unit consumption across the tenant or the 14-day granular item-level view.

73
MCQmedium

You are using the Microsoft Fabric Capacity Metrics app to monitor a Fabric capacity. You notice that the capacity frequently experiences throttling during peak hours, affecting interactive workloads. You need to reduce the likelihood of throttling without increasing capacity size. Which action should you take?

A.Reschedule background jobs to run during off-peak hours.
B.Set a workload-level cap on interactive operations.
C.Increase the capacity's max memory per query.
D.Enable autoscale for the capacity.
AnswerA

Background jobs consume capacity resources and can compete with interactive workloads during peak hours. By moving them to off-peak times, you reduce the overall load during peak periods, lowering the chance of throttling for interactive operations. This is a direct way to balance consumption without increasing capacity size, and it aligns with Fabric's workload management principles.

Why this answer

Moving background jobs to off-peak hours reduces competition with interactive workloads during peak times, directly lowering the chance of capacity throttling without increasing capacity size. Increasing memory per query, enabling autoscale, or capping interactive operations either do not address the root cause or violate the constraint of not increasing capacity.

Exam trap

The trap here is thinking that autoscale or per-query memory settings can solve throttling without increasing capacity, when the real solution is to shift background load away from peak periods.

74
MCQhard

You manage a Fabric capacity that runs both interactive workloads during business hours and scheduled background jobs overnight. Users report that interactive reports become slow in the late afternoon, and you want to confirm whether the capacity is being throttled and which operations are consuming the most capacity units. Which approach best meets this need?

A.Review the workspace's item lineage view to see which reports depend on which datasets and identify the heaviest chain.
B.Use the Microsoft Fabric Capacity Metrics app to review the multi-metric ribbon chart and timepoint detail for throttling and top consumers.
C.Check the Spark job history in the Monitoring hub to find long-running notebooks that overlap with business hours.
D.Open the Monitoring hub and sort activities by duration to find the slowest operations during the afternoon.
AnswerB

The Capacity Metrics app is purpose-built to show capacity utilization, throttling states, and which items or operations consume the most capacity units over time. Its ribbon chart and timepoint detail let you correlate the late-afternoon slowdown with throttling and identify the top consuming workloads, which is exactly what the scenario requires.

Why this answer

The Capacity Metrics app is the tool that expresses workload consumption in capacity units and surfaces throttling states over time, so it can both confirm throttling and rank the operations driving consumption. Activity and lineage views describe individual runs and dependencies but do not quantify capacity usage, and Spark job history covers only one workload type.

Exam trap

The trap here is using activity-level tools that show duration but not capacity units, when throttling can only be confirmed through capacity-level metrics.

75
MCQmedium

Refer to the exhibit. The refresh operation for a Power BI semantic model took 45 minutes to complete. What is the most efficient way to reduce the refresh time for this specific model?

A.Upgrade to a higher capacity SKU.
B.Configure incremental refresh.
C.Use Power BI Desktop to manually refresh.
D.Delete all historical data.
AnswerB

Incremental refresh allows Power BI to refresh only the partitions that have changed, rather than reloading the entire dataset. This is the industry-standard method for handling large models, as it minimizes the processing time and network traffic required for daily model updates.

Why this answer

Incremental refresh partitions the semantic model by a date/time column and only refreshes the most recent partition(s) plus any historical partitions that changed, instead of reprocessing the entire table. For a model that takes 45 minutes to fully refresh, this typically reduces refresh time to a fraction because only new or modified rows are loaded. It is the standard, most efficient optimization for large fact tables with time-based data.

Exam trap

The trap is assuming that throwing more hardware (higher SKU) at the problem is the 'efficient' fix, when the exam is testing whether you know incremental refresh is the architectural optimization for large, time-series semantic models.

How to eliminate wrong answers

Option A is wrong because upgrading the capacity SKU increases available compute and memory but does not reduce the volume of data being processed — the full refresh still scans and loads every row, so gains are marginal and costly. Option C is wrong because manually refreshing in Power BI Desktop does not change the amount of work performed and is not a scalable production solution. Option D is wrong because deleting historical data destroys analytical value and is not an optimization technique — it's data loss.

Page 1 of 2 · 102 questions totalNext →

Ready to test yourself?

Try a timed practice session using only Monitor Optimize Analytics Solution questions.