Blog   |   Cost Management   |   July 11, 2017

Ten Steps to Reducing Public Cloud Bills: Setting a Rightsizing Roadmap

Purchasing instances in the cloud is easier than ever, making teams more agile and efficient than was possible with on-premises infrastructure. With AWS cloud in particular, organizations have the option of leveraging a variety of EC2 instance types and families to meet specific capacity and performance requirements.
As Gartner highlighted in their recent rollup of Ten Moves to Lower Your AWS IaaS Costs, understanding the dynamics of EC2 families and types can help cloud-based organizations better manage and optimize their cloud expenses. In the next of our ten-part series, we add our take on their fourth recommendation: rightsize initially, baseline, rightsize again and repeat.

Baseline Before Buying

AWS EC2 instances offer different families and sizes to support a wide range of workloads. These instance families are typically best suited for certain types of activities—computing, memory, storage, or accelerated computing. Instance sizes, ranging incrementally from nano through 16xlarge, offer scalability and consistency.
When assessing needs to help inform their initial cloud instance purchases, an organization may first look to the CPU and memory of their current on-premise workload. As Gartner illustrates in their paper, matching current capacity to AWS options may seem easy, initially.
Yet this approach to purchasing cloud instances may lead to waste, if not thoroughly vetted.
As Gartner explains, getting an understanding of what is actually used or needed is crucial to making an informed EC2 purchase. On-premise sizes are based on what is provisioned—not necessarily what is utilized. Keeping this in mind can help your organization rightsize and baseline off the bat to mitigate under-utilization and wasted resources.
Additionally, considering instance family generations, as well as fixed performance versus burstable performance instances, can help your teams proactively navigate the highly dynamic nature of cloud costs.

A Workflow for Managing Instances

Deciding what to purchase to stretch your cloud spend takes considerable thought—but knowing how to manage it is where to true cost saving potential of the cloud comes into play.
CloudCheckr has outlined a workflow for continuously managing instances as you add more workloads and applications to the cloud:

  1. Inventory your RIs
  2. Identify mismatches of RIs with workloads
  3. Rebalance your RIs to eliminate mismatches
  4. Determine if you need to purchase additional RIs
  5. Review your savings

Although this specifies RIs, this approach aligns with Gartner’s methodology of rightsizing, baselining, then rightsizing again; as we know, in a rapidly-scaling cloud environment, leaving instances to manage themselves simply isn’t an option. For On-Demand or Spot Instances, the need to be proactive is even more important.
CloudCheckr offers many customizable reports, alerts, and best practice checks to help teams ensure their instances are fully utilized. With self-healing automation capabilities, organizations can be sure their instances are fully utilized while ensuring consistent performance for their workloads.
See CloudCheckr’s utilization capabilities in action: schedule a demo.
Read the previous post in this series, Ten Steps to Reducing Public Cloud Bills: Exploit Cloud Economics and keep going with part 5 on reserved instances.