Last month we wrote an article on our Rightsizing component which analyses and recommends a suitable instance type for your application based on standard CloudWatch metrics such as CPUUtlization, NetworkIn and NetworkOut. Today we announce support for capturing memory metrics from CloudWatch, enabling an even better recommendation.
While creating a cmpute job, the user gives us consent to capture and push memory metrics to CloudWatch. We do so by installing a CloudWatch agent on each instance provisioned via cmpute.
Like other metrics, cmpute analyses every instance’s memory consumption for a period of 15 days. cmpute identifies the peaks and valleys and calculates an average score. Based on this, cmpute then recommends a suitable instance type.
In our earlier blog post, we saw that cmpute suggests r3.large for an m4.xlarge resulting in 23% savings or an additional savings of 95 USD per instance per month. But taking memory metric into consideration, cmpute is now able to suggest c4.large for the same m4.xlarge resulting in 50% savings or an additional savings of 145 USD per instance per month.
Give it a try!
Please note that:
- Memory metric based recommendation is currently supported only for Linux instances
- The instance profile must have ‘CloudWatchFullAccess’ permission