This chapter covers AWS Compute Optimizer, a service that analyzes resource utilization and generates recommendations to reduce cost and improve performance. For the SOA-C02 exam, this topic appears in Domain 6: Cost Management (Objective 6.2: Implement cost optimization strategies). While not a heavy-weight topic, you can expect 1–2 questions that test your understanding of how Compute Optimizer works, its supported resources, recommendation types, and how to interpret findings. This chapter provides the depth needed to answer those questions confidently.
Jump to a section
Imagine you manage a fleet of delivery vans for a company. Each van has different engine specs, cargo loads, and routes. You want to reduce fuel costs without sacrificing delivery speed. A fuel economy advisor monitors each van's actual fuel consumption, engine RPM, idle time, and route terrain over several weeks. It then compares each van's performance against a baseline of similar vans (same model, similar cargo weight, similar route type). The advisor produces recommendations: 'Van #123 should be replaced with a smaller, more efficient model because its average fuel consumption is 30% higher than the fleet average for similar routes.' Or: 'Van #456 is underutilized — it only runs 2 hours per day; consider switching to a rental or electric scooter.' The advisor doesn't force changes; it provides evidence-based suggestions. If you ignore the advice, costs remain high. The advisor also tracks whether you implemented its suggestions and shows the projected savings. In AWS, Compute Optimizer works similarly: it analyzes your EC2 instances, Auto Scaling groups, and Lambda functions using historical utilization metrics (CPU, memory, network, storage) and compares them against known performance characteristics of instance families. It then generates recommendations to right-size (change instance type) or downsize (stop underutilized resources) to reduce cost without degrading performance. It uses machine learning models trained on millions of workloads to predict whether a change would improve performance or cost. The output is a set of findings: 'Over-provisioned,' 'Under-provisioned,' or 'Optimized.' You can then apply the recommendations manually or automate with AWS Systems Manager or Lambda. The key is that Compute Optimizer does not make changes — it advises, similar to how the fuel advisor does not drive the vans but provides data-driven suggestions.
What is AWS Compute Optimizer?
AWS Compute Optimizer is a machine learning-based service that analyzes the configuration and utilization of your AWS resources — specifically EC2 instances, Auto Scaling groups, Amazon EBS volumes, and Lambda functions — and generates recommendations to optimize costs and performance. It was launched in December 2019 and is one of the key tools in the AWS Cost Optimization pillar of the Well-Architected Framework.
The service works by collecting CloudWatch metrics (CPU, memory, network throughput, disk I/O) over a lookback period (default 14 days, but can be extended to 93 days with enhanced infrastructure metrics). It also considers instance type specifications, pricing, and workload patterns. Using machine learning models trained on millions of AWS workloads, it compares your resource's utilization against anonymized benchmarks of similar workloads. It then produces a finding: Over-provisioned, Under-provisioned, or Optimized. For over-provisioned resources, it suggests a smaller instance type or a different family that can handle the same load at lower cost. For under-provisioned resources, it suggests a larger instance type to prevent performance degradation.
How Compute Optimizer Works Internally
Compute Optimizer operates in three phases: data collection, analysis, and recommendation generation.
Data Collection Phase: - Compute Optimizer ingests CloudWatch metrics from the resources you opt in to analyze. For EC2 instances, it uses the following metrics (sampled every 5 minutes):
- CPUUtilization - NetworkIn / NetworkOut - DiskReadOps / DiskWriteOps - DiskReadBytes / DiskWriteBytes - For memory utilization, you must enable CloudWatch agent or a custom metric; otherwise, memory is not considered. Compute Optimizer can use enhanced infrastructure metrics if you enable detailed monitoring (1-minute metrics) for EC2 instances. - The lookback period is configurable: you can choose 14 days (default), 30 days, or 93 days. Longer lookback provides better recommendations for cyclical workloads. - Compute Optimizer also uses the instance type's specifications (vCPU, memory, EBS bandwidth, network bandwidth) and the current pricing (On-Demand, Reserved Instance, or Savings Plan) to calculate potential savings.
Analysis Phase: - The service compares your resource's utilization percentiles (P50, P90, P99) against the capacity constraints of the current instance type. For example, if CPU utilization is consistently below 40% at P99, the instance is likely over-provisioned. If CPU utilization frequently spikes to 90% or more, it may be under-provisioned. - Compute Optimizer uses a machine learning model that considers workload characteristics such as burstiness, periodicity, and correlation between metrics. It does not simply compare averages; it uses percentile thresholds. - The model also evaluates whether a different instance family (e.g., moving from M5 to T3) or a different architecture (e.g., ARM-based Graviton) would be beneficial.
Recommendation Generation Phase: - Compute Optimizer produces findings and recommendations. Findings are one of: - Over-provisioned: The resource is larger than needed. Recommendation includes a smaller instance type or a different family. - Under-provisioned: The resource is too small, risking performance issues. Recommendation includes a larger instance type. - Optimized: The resource is appropriately sized. - Not optimized (for EBS volumes): The volume is not meeting performance needs or is over-provisioned. - For each recommendation, Compute Optimizer provides:
Current instance type and configuration.
Recommended instance type and configuration.
Estimated monthly savings (in USD).
Performance risk score (Low, Medium, High) indicating the likelihood that the new instance will meet workload demands.
Reason codes explaining the recommendation (e.g., CPUOverProvisioned, MemoryUnderProvisioned).
Key Components, Values, Defaults, and Timers
Supported Resources: EC2 instances (including Spot and On-Demand), Auto Scaling groups, EBS volumes (gp2, gp3, io1, io2), and Lambda functions.
Lookback Period: Default is 14 days. Can be set to 30 or 93 days. The longer the lookback, the more accurate the recommendation for cyclical workloads.
Enhanced Infrastructure Metrics: If you enable detailed monitoring (1-minute metrics) for EC2, Compute Optimizer can use more granular data.
Memory Utilization: Not collected by default. Must be enabled via CloudWatch agent or custom metric. If not enabled, Compute Optimizer cannot make memory-based recommendations.
Recommendation Preferences: You can set preferences to exclude certain instance families (e.g., exclude T2 instances) or to only recommend instances from a specific generation (e.g., current generation only).
Opt-in: Compute Optimizer is opt-in per account. You must enable it for each resource type (EC2, Auto Scaling, EBS, Lambda) via the console or API.
Savings Calculation: Uses On-Demand pricing by default, but you can provide your effective hourly rate if you have Reserved Instances or Savings Plans.
Performance Risk: Low (probability < 5% of performance degradation), Medium (5-20%), High (>20%).
Recommendation Export: You can export recommendations to S3 in CSV or JSON format.
Automation: You can automate applying recommendations using AWS Systems Manager Automation (runbooks) or by triggering Lambda functions via EventBridge events when new recommendations are generated.
Configuration and Verification Commands
To enable Compute Optimizer for EC2 instances using AWS CLI:
aws compute-optimizer update-enrollment-status --status Active --member-account-id 123456789012To get recommendations for EC2 instances:
aws compute-optimizer get-ec2-instance-recommendations --instance-arns arn:aws:ec2:us-east-1:123456789012:instance/i-0abcd1234To list all findings for an account:
aws compute-optimizer get-recommendation-summariesTo set recommendation preferences (e.g., exclude T2 instances):
aws compute-optimizer put-recommendation-preferences --resource-type EC2 --scope Instance --excluded-families t2To export recommendations to S3:
aws compute-optimizer export-ec2-instance-recommendations --s3-destination-config bucket=my-bucket,keyPrefix=compute-optimizer/To verify if Compute Optimizer is enabled:
aws compute-optimizer get-enrollment-statusHow Compute Optimizer Interacts with Related Technologies
CloudWatch: Compute Optimizer relies on CloudWatch metrics. Without sufficient metric history (minimum 14 days), it cannot generate recommendations. If you enable detailed monitoring, it uses 1-minute data instead of 5-minute.
AWS Trusted Advisor: Trusted Advisor also provides cost optimization checks (e.g., idle instances, underutilized EBS volumes). However, Trusted Advisor is rule-based and does not use machine learning. Compute Optimizer is more sophisticated and provides instance family changes, not just downsizing.
AWS Cost Explorer: Cost Explorer provides cost and usage data but does not provide resource optimization recommendations. Compute Optimizer integrates with Cost Explorer to show potential savings.
AWS Systems Manager: You can use Systems Manager Automation runbooks to automatically apply Compute Optimizer recommendations (e.g., stop an instance, change instance type, start it).
AWS Lambda: You can set up EventBridge rules to trigger Lambda functions when new recommendations are generated, enabling custom workflows (e.g., send notification to Slack, automatically apply if risk is low).
AWS Organizations: Compute Optimizer can be enabled for all accounts in an organization using a management account. You can view recommendations across accounts.
Supported Resources and Their Specifics
EC2 Instances: - Compute Optimizer analyzes all running instances that have been active for at least 30 hours over the lookback period. - It considers CPU, memory (if available), network, and EBS metrics. - Recommendations include changing instance type, family, or size. It may also recommend moving to a different purchase option (e.g., from On-Demand to Reserved Instance) if applicable, but the primary recommendation is instance type change.
Auto Scaling Groups: - Compute Optimizer analyzes the launch template or configuration of the Auto Scaling group and the utilization of instances in the group. - It recommends changes to the instance type in the launch template/configuration. - It also considers the group's scaling policies and metrics.
EBS Volumes: - Compute Optimizer analyzes gp2, gp3, io1, and io2 volumes. - It looks at IOPS and throughput utilization. - Findings: Over-provisioned (volume size or performance is too high), Under-provisioned (volume is bottleneck), Optimized. - Recommendations include volume type changes (e.g., gp2 to gp3) and size adjustments.
Lambda Functions: - Compute Optimizer analyzes functions that have been invoked at least 50 times in the lookback period. - It considers duration, memory size, and number of invocations. - Recommendations include memory size changes (e.g., from 128 MB to 256 MB) to reduce cost or improve performance.
Exam-Relevant Details
Compute Optimizer does NOT automatically implement recommendations. It only provides suggestions. You must apply them manually or via automation.
The default lookback period is 14 days. You can change it to 30 or 93 days.
Memory utilization is not automatically collected. You must set up the CloudWatch agent or push custom metrics. If you don't, Compute Optimizer will still generate recommendations based on CPU, network, and disk, but memory-based findings will be missing.
Compute Optimizer supports both instance families and sizes. It may recommend moving from a general-purpose instance (M5) to a compute-optimized instance (C5) if the workload is CPU-bound.
Performance risk is a key metric: Low risk means high confidence the recommendation will not cause performance issues. High risk means caution is advised.
Compute Optimizer is free of charge. There is no additional cost for using the service.
It is available in all commercial AWS regions and AWS GovCloud (US).
You can exclude specific instance families from recommendations (e.g., exclude T2 instances because they are burstable and may not be suitable for production).
Compute Optimizer uses machine learning models that are continuously updated. The models are trained on anonymized AWS customer data.
For EC2 instances, Compute Optimizer can recommend moving to an ARM-based Graviton instance (e.g., from M5 to M6g) if the workload is compatible. This can result in significant cost savings.
The service integrates with AWS Compute Optimizer API, AWS CLI, AWS SDK, and AWS Management Console.
Common Exam Traps
Trap: Compute Optimizer automatically resizes instances. Reality: It only generates recommendations. You must apply them.
Trap: Memory utilization is always analyzed. Reality: Only if you enable memory metrics via CloudWatch agent.
Trap: Compute Optimizer works with all EBS volume types. Reality: It supports gp2, gp3, io1, io2. It does not support st1, sc1, or standard (magnetic) volumes.
Trap: You need to enable Compute Optimizer for each resource type separately. Reality: Correct. You opt in for EC2, Auto Scaling, EBS, and Lambda individually.
Trap: The lookback period is fixed at 14 days. Reality: You can change it to 30 or 93 days.
Trap: Compute Optimizer can only recommend smaller instances. Reality: It also recommends larger instances if the resource is under-provisioned.
Trap: Compute Optimizer is a paid service. Reality: It is free.
Best Practices
Enable Compute Optimizer for all account types to get a comprehensive view of optimization opportunities.
Enable detailed monitoring (1-minute metrics) for EC2 instances to get more accurate recommendations.
Set up memory metrics for EC2 instances to enable memory-based recommendations.
Review recommendations regularly (e.g., weekly) and apply them during maintenance windows.
Use preference settings to exclude instance families that are not suitable for your workload (e.g., burstable instances for steady-state workloads).
Automate the application of low-risk recommendations using Systems Manager Automation or Lambda.
Export recommendations to S3 for historical analysis and reporting.
Use Compute Optimizer in conjunction with AWS Cost Explorer and Trusted Advisor for a holistic cost optimization strategy.
Enable Compute Optimizer
Navigate to the AWS Compute Optimizer console. Click 'Get started' or use the API to opt in. You must enable Compute Optimizer for each resource type separately: EC2 instances, Auto Scaling groups, EBS volumes, and Lambda functions. You can also enable it across an AWS Organization using the management account. Once enabled, Compute Optimizer begins collecting CloudWatch metrics for the selected resources. There is no cost for enabling the service. The opt-in status can be verified via the console or CLI command: `aws compute-optimizer get-enrollment-status`.
Configure Lookback Period and Preferences
Set the lookback period (14, 30, or 93 days) via the console or CLI. Longer periods improve recommendation accuracy for cyclical workloads. Configure preferences such as excluded instance families (e.g., exclude T2 or previous generation instances) and whether to include enhanced infrastructure metrics (requires detailed monitoring). You can also specify your effective hourly rate if you have Reserved Instances or Savings Plans to get accurate savings estimates. Preferences are set per resource type using the `put-recommendation-preferences` API.
Enable Memory Metrics (Optional but Recommended)
By default, Compute Optimizer does not collect memory utilization. To enable memory-based recommendations for EC2 instances, install the CloudWatch agent on the instance and configure it to report memory metrics (e.g., mem_used_percent). Alternatively, you can push custom metrics via the PutMetricData API. Without memory metrics, Compute Optimizer can still generate recommendations based on CPU, network, and disk, but it may miss memory over-provisioning or under-provisioning. This step is crucial for accurate recommendations for memory-intensive workloads.
Review Findings and Recommendations
After at least 14 days of data collection, Compute Optimizer generates findings. Access the console to view a summary dashboard showing the number of over-provisioned, under-provisioned, and optimized resources. For each resource, click to see detailed recommendations: current configuration, recommended configuration, estimated monthly savings, performance risk, and reason codes. You can filter by resource type, region, account, and finding. Use the CLI command `get-ec2-instance-recommendations` to retrieve recommendations programmatically. Export recommendations to S3 for further analysis.
Apply Recommendations
Compute Optimizer does not apply recommendations automatically. To apply a recommendation for an EC2 instance, you can stop the instance, change the instance type, and start it again. For Auto Scaling groups, update the launch template or configuration. For EBS volumes, modify the volume type or size. For Lambda functions, update the memory configuration. To automate, use AWS Systems Manager Automation runbooks that perform these steps. You can also set up EventBridge rules to trigger Lambda functions when new recommendations are generated, applying only low-risk recommendations automatically. Always test recommendations in a non-production environment first to avoid performance impact.
Enterprise Scenario 1: E-commerce Platform with Cyclical Traffic
A large e-commerce company runs hundreds of EC2 instances across multiple Auto Scaling groups to handle web traffic. Traffic spikes during holiday seasons and flash sales. The cloud engineering team uses Compute Optimizer to identify over-provisioned instances during off-peak periods. They set the lookback period to 93 days to capture both peak and low seasons. Compute Optimizer recommends downsizing many m5.large instances to m5.xlarge? Actually, downsizing from m5.large to t3.medium for low-traffic periods? Wait, the recommendation is to move from m5.xlarge to m5.large for instances that rarely exceed 20% CPU. The team creates a Systems Manager Automation runbook that applies low-risk recommendations during maintenance windows. They also exclude T2 instances to avoid burstable performance issues in production. After three months, they achieve 35% cost savings without any performance degradation. The key is that they enabled memory metrics via CloudWatch agent, so Compute Optimizer could accurately assess memory utilization.
Enterprise Scenario 2: Financial Services with Compliance Requirements
A bank runs a mix of On-Demand and Reserved Instances for its trading applications. They need to maintain high performance and cannot tolerate latency. Compute Optimizer is used to identify under-provisioned instances that could cause performance bottlenecks. The bank configures Compute Optimizer to use enhanced infrastructure metrics (1-minute CloudWatch) for critical instances. Compute Optimizer identifies several instances that are under-provisioned on memory (high swap usage). The team increases memory by moving from m5.large to m5.xlarge, which resolves swap issues and improves application response time. They also use Compute Optimizer to validate that their Reserved Instance purchases are optimal. The service shows that some Reserved Instances are for instance types that are over-provisioned, leading to waste. They adjust future purchases accordingly.
Enterprise Scenario 3: Serverless Application with Lambda Functions
A SaaS company uses AWS Lambda for its backend microservices. They have hundreds of functions with varying memory settings (128 MB to 3 GB). Compute Optimizer analyzes invocation patterns and durations. It recommends increasing memory for functions that are CPU-bound (e.g., from 128 MB to 256 MB) to reduce duration and overall cost (since cost is a function of memory * duration). For other functions, it recommends decreasing memory if utilization is low. The company automates the application of recommendations using a Lambda function triggered by EventBridge. They set a rule to only apply recommendations with low performance risk. Over six months, they reduce Lambda costs by 20% while maintaining performance.
Common Pitfalls in Production
Not enabling memory metrics: Many teams skip this, leading to incomplete recommendations. Memory is often the bottleneck.
Ignoring performance risk: Applying high-risk recommendations without testing can cause outages.
Using default 14-day lookback for cyclical workloads: This misses seasonal patterns. Use 93 days for annual cycles.
Not excluding unsuitable families: For example, T2 instances are burstable and may not be suitable for steady-state workloads. Exclude them via preferences.
Manual application at scale: For large fleets, manual application is error-prone. Automate with Systems Manager or Lambda.
Over-reliance on Compute Optimizer alone: Combine with Trusted Advisor for idle resources and Cost Explorer for savings plans.
What SOA-C02 Tests on Compute Optimizer
Domain 6: Cost Management (Objective 6.2: Implement cost optimization strategies). The exam tests your understanding of:
The purpose and capabilities of Compute Optimizer.
Supported resources (EC2, Auto Scaling, EBS, Lambda).
How recommendations are generated (lookback period, metrics used).
How to interpret findings (Over-provisioned, Under-provisioned, Optimized).
The importance of memory metrics.
The fact that Compute Optimizer does not auto-apply recommendations.
Performance risk and savings estimates.
Integration with other services (CloudWatch, Systems Manager, Lambda).
Common Wrong Answers and Why Candidates Choose Them
'Compute Optimizer automatically resizes instances to reduce cost.' Candidates confuse Compute Optimizer with AWS Auto Scaling or AWS Instance Scheduler. Reality: Compute Optimizer only recommends; you must apply changes.
'Compute Optimizer requires no configuration — it works out of the box for all resources.' Candidates think it's like Trusted Advisor. Reality: You must opt in for each resource type and enable memory metrics manually.
'Compute Optimizer uses a 30-day lookback period by default.' The default is 14 days. Candidates often misremember this because many AWS services use 30-day defaults.
'Compute Optimizer can recommend Reserved Instances.' It recommends instance type changes, not purchase options. However, it can show savings if you provide your effective rate.
'Compute Optimizer analyzes memory utilization by default.' It does not. Memory metrics require CloudWatch agent installation.
Specific Numbers and Terms on the Exam
Lookback period: 14, 30, 93 days (default 14).
Supported resources: EC2, Auto Scaling groups, EBS (gp2, gp3, io1, io2), Lambda.
Findings: Over-provisioned, Under-provisioned, Optimized.
Performance risk: Low, Medium, High.
Reason codes: CPUOverProvisioned, MemoryUnderProvisioned, etc.
Minimum data: 30 hours of activity over lookback period for EC2; 50 invocations for Lambda.
Enhanced infrastructure metrics: requires detailed monitoring (1-minute).
Compute Optimizer is free.
Edge Cases and Exceptions
If an instance has been running less than 30 hours, no recommendations are generated.
For Lambda, functions with fewer than 50 invocations in the lookback period are not analyzed.
EBS volumes must be in use (attached to an instance) to be analyzed.
Compute Optimizer does not support EBS volumes of type st1, sc1, or standard (magnetic).
If you disable Compute Optimizer, all historical data is deleted after 90 days.
When using AWS Organizations, the management account can view recommendations for member accounts only if those accounts have opted in.
How to Eliminate Wrong Answers
If a question states that Compute Optimizer automatically applies changes, it's false.
If a question mentions memory analysis without CloudWatch agent, it's likely a distractor.
If a question gives a lookback period other than 14, 30, or 93, it's wrong.
If a question says Compute Optimizer is paid, it's wrong.
If a question says it supports all EBS volume types, look for the exception (magnetic, st1, sc1).
Use the mechanism: Compute Optimizer is a recommendation engine, not an automation tool. It relies on CloudWatch metrics and machine learning.
Compute Optimizer is a free service that generates right-sizing recommendations for EC2, Auto Scaling, EBS, and Lambda using machine learning.
Default lookback period is 14 days; can be set to 30 or 93 days.
Memory utilization is not collected by default; requires CloudWatch agent.
Compute Optimizer does NOT automatically apply recommendations; you must implement changes manually or via automation.
Findings: Over-provisioned, Under-provisioned, Optimized.
Performance risk: Low, Medium, High.
Supports EBS volume types: gp2, gp3, io1, io2. Does not support st1, sc1, or magnetic.
Minimum 30 hours of activity for EC2; 50 invocations for Lambda to generate recommendations.
You can exclude specific instance families (e.g., T2) via preferences.
Integrates with CloudWatch, Systems Manager, Lambda, and EventBridge.
These come up on the exam all the time. Here's how to tell them apart.
AWS Compute Optimizer
Uses machine learning to analyze utilization patterns.
Supports EC2, Auto Scaling, EBS, and Lambda.
Provides instance type recommendations (right-sizing).
Requires opt-in per resource type.
Free of charge.
AWS Trusted Advisor (Cost Optimization)
Rule-based checks (e.g., idle instances, underutilized volumes).
Supports EC2, EBS, and other services but not Lambda for cost.
Checks for idle instances and underutilized EBS volumes (not instance type changes).
Automatically enabled for AWS accounts.
Free for basic checks; Business and Enterprise support plans unlock more checks.
Mistake
Compute Optimizer automatically applies recommendations to optimize resources.
Correct
Compute Optimizer only generates recommendations. You must manually apply them or use automation (e.g., Systems Manager Automation or Lambda). It does not make any changes to your resources.
Mistake
Compute Optimizer requires no additional setup — it works immediately for all resources.
Correct
You must opt in to Compute Optimizer for each resource type (EC2, Auto Scaling, EBS, Lambda). Additionally, memory metrics are not collected by default; you need to install the CloudWatch agent to enable memory-based recommendations.
Mistake
Compute Optimizer uses a 30-day lookback period by default.
Correct
The default lookback period is 14 days. You can change it to 30 or 93 days for better accuracy, especially for cyclical workloads.
Mistake
Compute Optimizer can recommend Reserved Instance purchases.
Correct
Compute Optimizer recommends instance type changes (right-sizing), not purchase options like Reserved Instances or Savings Plans. However, it can estimate savings based on your effective hourly rate if you provide it.
Mistake
Compute Optimizer analyzes memory utilization for all EC2 instances automatically.
Correct
Memory utilization is not collected by default. You must enable the CloudWatch agent on your instances to report memory metrics. Without it, Compute Optimizer cannot make memory-based recommendations.
Reveal each answer, then mark whether you got it right. Score 60%+ to unlock the next chapter.
No, Compute Optimizer only generates recommendations. It does not automatically make any changes to your resources. You must manually apply the recommendations or use automation tools like AWS Systems Manager Automation or Lambda functions triggered by EventBridge.
Compute Optimizer uses CloudWatch metrics: CPUUtilization, NetworkIn, NetworkOut, DiskReadOps, DiskWriteOps, DiskReadBytes, DiskWriteBytes. Memory utilization is only considered if you enable the CloudWatch agent to report memory metrics. Without memory metrics, recommendations are based on CPU, network, and disk only.
Compute Optimizer requires at least 14 days of utilization data (default lookback period) to generate recommendations. If you change the lookback period to 30 or 93 days, it will take that long to collect sufficient data. For new resources, you must wait until the lookback period is filled.
Yes, if your workload is compatible, Compute Optimizer may recommend moving to ARM-based Graviton instances (e.g., from M5 to M6g). These recommendations can significantly reduce cost. However, you should test compatibility before applying.
Compute Optimizer is available in all commercial AWS regions and AWS GovCloud (US). It is not available in China regions or other isolated regions at this time.
Performance risk is the estimated probability that the recommended instance type will not meet your workload's performance requirements. Low risk (<5%) means high confidence. Medium risk (5-20%) means moderate confidence. High risk (>20%) means you should test carefully before applying.
Yes, you can set recommendation preferences to exclude specific instance families (e.g., T2, T3, previous generation instances). This is useful if you want to avoid burstable instances for steady-state workloads or if you have compliance requirements.
You've just covered AWS Compute Optimizer — now see how well it sticks with free SOA-C02 practice questions. Full explanations included, no account needed.
Done with this chapter?