The EC2 Right Sizing tool analyses all the instances in your account with a max CPU utilization lower than 50% plus max Memory utilization lower than 50% and determines a more cost-effective instance type for that workload, if available. The RightSizing results are populated periodically ( currently at 15 days )
Initial Signup
Once the signup is successfully completed, you will see the Right Sizing results in the next day under 'Manage - Right Sizing' section.
You can apply the filter to see the results based on considering the memory metrics or not. If memory metrics is ignored, it will not take the memory into consideration when choosing the appropriate type.
There is a option to choose the timeline how often to run right sizing solution - 15 or 30 days. Also, an option to be run for all the regions or only the region selected.
Last run field indicates the date when these data was captured and the next run date indicates when the next run will be happening.
EC2 Rightsizing Recommendations
This shows the instances with less than 50% max CPU utilization for the given period ( either 15 or 30 days) that are the likely candidates for Right Sizing.
Using memory metrics will improve the accuracy of the recommendations significantly. If the instances do not have Cloud watch agent installed, it is highly recommended that it be installed. The instructions to install Cloud watch agent can be found here.
Potential Spend
This indicates the potential spend of the currently running EC2 instances for the month
Maximum Savings per month
This is a potential savings possible if all the recommendations are implemented. If there are multiple recommendations, the recommendation with the highest savings is considered for the calculation.
The instance with the current configuration and max CPU utilization and Max. network will be displayed. If CloudWatch agent metrics are available , then Max. memory metrics are also displayed.
Each instance will have up to 5 recommendations. To view the recommendations, click on 'Recommendations'
Each recommendation will display the new CPU/Memory/ Network and the reason for it to be recommended. The reasons can be CPU, Network, Memory, Processor Type ( Intel based, AMD based, ARM based). Ideally, choosing a recommendation with all three reasons is advisable. In some cases, one of the reasons may not be a match. These should be evaluated in the context of the applications running on the instance if that particular attribute is required or not.
For e.g. m4.xlarge with 10% sustained CPU and 40% burst of 10 minutes with 3.5 GB of memory usage and low memory usage. In this case, M4.large as being most compatible with the current usage by virtue of measured metrics. However, M5.large may be a good fit if the ENA drivers are installed. m5a.large may provide slightly better savings if performance variance of 5-10% is acceptable and applications are compatible with AMD processors. Based on the sustained CPU burst , moving to T3 series can be a good option. These will be provided in the recommendation with Reasons and Compatibility between instance types.
Reason
Displays the reasons for making the recommendations in terms of having similar characteristics as the existing instance.
Compatibility
Displays the attributes where the new instance type is NOT same as the earlier instance type. These are characteristics that are typically based on applications running on the instances. Understanding those details will help make better choices.
Processor - Intel/AMD
e.g M5a , T3a and other 'a' type instances are AMD based and can offer 10% additional savings over Intel based T3, M5 and other 5 family based instances. But it is suitable for workloads that are AMD compatible and are suited for general workloads.
https://aws.amazon.com/ec2/amd/
Nitro
The Nitro based EC2 instances represent the next generation of instance types from AWS. They provide better performance and cost less by upto 20% than existing 4 family series of similar configuration. You can find more details at https://aws.amazon.com/blogs/aws/amazon-ec2-update-additional-instance-types-nitro-system-and-cpu-options/
However, before upgrading to newer family of instance types, the drivers need to be upgraded for the AMI if the AMIs also predate the release of Nitro ( Oct 2018).
32-bit / 64-bit
Some instance types support both 32/64 bit applications e.g. T2 but the newer T3 support only 64 bit applications. If you have applications that can only run 32 bit applications, then upgrading the instance types that support only 64-bit apps is not suitable.
Burstable/ Provisioned:
The T2/T3 series are suitable for workloads that have low CPU/network utilization but need occasional 'burst' of CPU. If your overall usage is less, consider converting to T2/T3 instance types.
Status
The status of the recommendation. The states can be pending, In Progress, Completed, Failed or Reverted. The initial state will be 'Pending'.
Action
Please ensure the pre requisites are met before updating an Instance. If a recommendation is to be implemented, click on 'Update' action . This will bring up the confirmation dialog to initiate the right size process in the background. The periodic status updates will be updated in the 'Status' column by manually refreshing the page. Once the update is completed, the status for the selected recommendation will display 'Completed' and other recommendations to ' NA'.
There will be downtime for the instance while the instance type change is in progress.
Reverting to original instance type.
In some rare cases, you might want to revert back to original configuration. This is possible with the 'Revert' action available once the right sizing update is completed successfully. In case, other recommendation needs to be applied, choose 'Override' option. This will apply the new recommendation.
Cost Implications
For actions that are reverted, there is no cost savings claimed. For recommendations that override, the new cost savings will be considered instead of the cost savings for the initially accepted recommendation. The cost savings is calculated as the difference is cost for 30 days between old and new instance type.
The Right size recommendations will run automatically every 15 days.