DP-900 Describe an analytics workload on Azure Practice Question
A financial services company is building a real-time fraud detection system. Transactions are streamed from multiple sources into Azure Event Hubs. The system must run a trained machine learning model (scored in near real-time) to flag suspicious transactions. The model is a Python pickle file that needs to be deployed as a web service with low latency (under 100 ms per prediction). The data engineering team wants to use a serverless compute option to run the scoring logic, and the solution must integrate with Azure Stream Analytics for alerting. Which Azure service should you use to deploy the model?
⚠ Common exam trap
Test-takers frequently choose Azure Functions because it is serverless and familiar, but they overlook the strict latency requirement (under 100 ms) and the need for native integration with Azure Stream Analytics, which Azure Machine Learning managed online endpoints satisfy directly.
Answer choices
Why each option matters
Answer the question above first, then reveal the full breakdown to understand why each option is right or wrong.
Correct answer & explanation
✓
Azure Machine Learning managed online endpoint
Azure Machine Learning managed online endpoints are the correct choice because they are designed for deploying trained models (including Python pickle files) as low-latency web services (under 100 ms per prediction) with serverless compute. They natively integrate with Azure Stream Analytics for alerting, allowing real-time scoring of streaming transactions from Event Hubs without managing infrastructure.
Answer analysis
Option-by-option breakdown
For each option: why learners choose it and why it is or isn't the right answer here.
- ✗
Azure Functions
Why it's wrong here
Azure Functions can technically host a trained model by wrapping it in a function, but this serverless compute service is not optimized for machine learning inference. Because Functions scale to zero during idle periods, incoming requests may encounter cold-start delays that add multiple seconds of latency, which is unacceptable for real-time fraud detection. Additionally, Functions provide limited control over concurrency and resource allocation, making it difficult to guarantee the low and predictable response times required for transaction screening.
- ✓
Azure Machine Learning managed online endpoint
Why this is correct
Azure Machine Learning managed online endpoints are purpose-built for deploying models as production-grade, real-time REST APIs. These endpoints handle the underlying infrastructure, including load balancing and auto-scaling, so you get a serverless experience with low latency and high availability. For fraud detection, the endpoint can be invoked from Azure Stream Analytics or any consumer over HTTP, returning predictions in milliseconds and thus meeting the strict sub-second performance requirements of real-time transaction monitoring.
- ✗
Azure Kubernetes Service (AKS)
Why it's wrong here
Azure Kubernetes Service focuses on container orchestration rather than machine learning serving, so you must handle cluster provisioning, node pool scaling, rolling updates, and ingress configuration yourself. While AKS can achieve low latency with a well-tuned deployment, it is not a serverless offering and demands ongoing operational expertise to maintain the environment for real-time inference. Compared to a managed online endpoint, AKS introduces significant complexity in integrating with Azure Machine Learning and monitoring model health.
- ✗
Azure Databricks
Why it's wrong here
Azure Databricks is an open analytics platform centered on Apache Spark, which is optimized for large-scale batch data processing and iterative model training rather than serving individual prediction requests. To use Databricks for real-time inference you would have to create a streaming workload or a custom REST proxy on an interactive cluster, but that design still carries high per-request latency and lacks native integration with Azure Stream Analytics. For fraud detection needing immediate responses per transaction, Databricks workflow is better suited to periodic scoring of historical data than to synchronous API calls.
Quick reference
Cloud Service Model Comparison
| Model | You Manage | Provider Manages | Examples |
|---|---|---|---|
| IaaS | OS, runtime, apps, data | Hardware, hypervisor, networking | EC2, Azure VMs, GCP Compute Engine |
| PaaS | Apps and data | OS, runtime, middleware, hardware | Elastic Beanstalk, Azure App Service |
| SaaS | Data and settings only | Everything else | Microsoft 365, Salesforce, Workday |
| FaaS / Serverless | Function code only | Infra, scaling, runtime | Lambda, Azure Functions, Cloud Run |
| CaaS | Containers and apps | Kubernetes, OS, hardware | EKS, AKS, GKE |
Go deeper
Related to this question
Learn chapter
Data Roles and Core Concepts
Key term
Azure Stream Analytics
Azure Stream Analytics is a fully managed, real-time data processing service that analyzes and transforms high volumes of streaming data from various sources to deliver low-latency insights and trigger actions.
Key term
Service
A service is a software component or system that performs a specific function and is available to be used by other programs or users over a network.
About these practice questions
One of 820 original DP-900 practice questions on Courseiva, each with a full explanation and wrong-answer analysis — not exam dumps or protected exam content. Learn why practice questions differ from exam dumps →
JA
Written by Johnson Ajibi, MSc IT Security
Senior Network & Security Engineer · founder of Courseiva
This DP-900 practice question is part of Courseiva's free Microsoft certification practice question bank. Courseiva provides original exam-style practice questions with explanations, topic-based practice, mock exams, readiness tracking, and study analytics to help learners prepare for the DP-900 exam.