AWS Cost optimization

A large Enterprise company with multiple accounts all spending over 1.8M per month wants to cut expenses for itself and its customers. They use both reserved instances and savings plans within AWS.


Original Monthly Expenses

2 d

Time spent


Savings Unused resources discovered. 


Savings Over-provisioned EKS resources.


Savings Optimising Disk


Total Savings 

Use case

Company’s AWS users tag their resources, which enables allocating cost grouped by tag.  When applying it together with pricing info and utilization analytics they get an opportunity to highlight unoptimized cloud usage and scale down instances which cause it.  This use case helps the organization to balance workload to reduce budget waste caused by over provisioned resources.


  • Calculate product monthly cost
  • Detect underutilized resources
  • Scale according to utilization
  • Calculate savings


  • Use Optimization lab to allocate product’s cost and filter associated resources with low-loaded CPU
  • Update Auto Scaling groups, cluster configurations, service configurations and EC2 Instance with updated resource type to better match the needed CPU but also ensure memory requirements are still met.
  • After collecting resources’ CPU utilization data  for a 90 day period along with their tags and cost, the task is to detect overprovisioned ones that have <10% of CPU load at their peaks. Then define applicable replacement to meet load requirements and set up an appropriate AutoScaling group.
  1. CloudAvocado Exposed the resources in question, which then allowed the team to focus on what to do with the resources as opposed to trying to find these resources.
  2. Consolidate Resources.
  3. Reconfigure Auto-scaling Groups
  4. Downgrade over provisioned resources based on CPU but keep memory the same or higher.
  5. Upgrade resources for reduced cost.

Whats next?

Easy way to discover more about platform capabilities

Scroll to Top

Contact Us

Sign up so we can tailor your support experience. If that’s not possible, we’d still like to hear from you.