New Year’s Resolution – Reduce AWS EC2 costs by 85%

Paddy Raghavan | 05th Jan, 2017

We always hear about the same-old boring New Year’s resolutions that we end up breaking very soon. If you are an AWS user, there is an easy way to have (and stick with it) a New Year’s resolution of reducing your AWS costs (by up to 85%). Read on to find how to do it.

AWS Enterprise Adoption

Image Courtesy: AWS Enterprise Customers (Sample List)

Enterprises are adopting AWS at a never-before-seen pace. The factors leading to this, but not limited to, are:

  • AWS takes security very seriously
    • It boasts having one of the largest talent pool of security engineers working alongside the development teams. They work round-the-clock to ensure that security is never compromised.
  • AWS Certifications
    • AWS provides training and certifications to enable your workforce to learn Cloud Computing skills.
  • AWS Marketplace
    • AWS has an active marketplace incorporating solutions from Microsoft, Oracle, and other system integrators and software vendors. This makes it easy to cherry-pick the applications you need and get started immediately.

The workflows or applications commonly used in most enterprises that heavily use AWS EC2 are: Auto Scaling Groups, BeanStalk and ELB (for Web applications), Elastic MapReduce (for Analytics) and CI / CD / Configuration Management like Chef, Puppet, Terraform and Jenkins.

Challenges

AWS is all about scale. It offers 16 regions and 42 Availability Zones worldwide. In each region, there are 50+ instance types, 4 instance lifecycles; On-Demand, Reserved, Spot or Spot Blocks. Then there are instance categories – m3, c3, i2, r3 to name a few with varying instance sizes: large, xlarge, 2xlarge, 4xlarge, and so on. Some of these are memory-optimized, some compute-optimized while others are general-purpose. Add to this, choosing an optimum bid price for launching Spot Instances depends on various factors such as region, Availability Zone, Spot market price fluctuations, and time of day. There are 600+ variations on setting a bid price. The sheer number of choices from AWS also brings in the complexity.

The immediate question on your mind is – “I want to use AWS in the most efficient and cost-effective way. How do I do it without breaking the bank or my application?”

CMPUTE

CMPUTE is a SaaS solution that balances EC2 workloads over on-demand, spot, and reserved instances to achieve up to 85% savings on your EC2 costs without compromising on your application availability / uptime.

Use cases where cmpute can seamlessly fit in and bring the 80% savings:

  • Web-apps on Auto Scaling Groups behind Elastic Load Balancer
    • CMPUTE reads the ASG information and starts managing it for you. It provides a unique “Instance Distribution Model” approach where:
      • min instance count will be launched as OnDemand Instances
      • desired instance count will be maintained using Spot Instances
      • max instance count will be launched on Spot Instances during scale-out
    • This ensures that the application uptime is never compromised while assuring of maximum cost savings through the usage of Spot Instances.
  • Elastic MapReduce analytical jobs
    • CMPUTE runs your analytical workloads using TASK nodes on Spot Instances. It does not touch MASTER and CORE nodes thereby ensuring data consistency at all times.
    • CMPUTE autoscales TASK nodes on Spot Instances so that the job runs to completion successfully at the earliest.
    • It finally terminates the TASK nodes so that there are no additional costs.
    • This ensures that the analytical job runs quickest and at the most cost-efficient way.
  • Worker Nodes on Auto Scaling Groups
    • CMPUTE reads the ASG information and starts managing it for you. It provides a unique “Instance Distribution Model” approach where:
      • min instance count will be launched as OnDemand Instances
      • desired instance count will be maintained using Spot Instances
      • max instance count will be launched on Spot Instances during scale-out
    • This ensures that a skeletal pool of On-Demand Instances caters to the incoming jobs while the traffic spikes can be handled using Spot Instances so that you get maximum savings.
  • Continuous Integration / Continuous Deployment
    • CMPUTE supports CI / CD deployments, be it Push-based or Pull-based. It easily fits into your existing CodeDeploy, Chef or Puppet pipeline. Using our “Custom Tags” feature, CMPUTE will be able to deploy new versions of your applications into all EC2 instances.
    • An Auto Scaling Group linked to a CI pipeline, say CodeDeploy, will benefit from the earlier “Instance Distribution Model” approach to give you maximum savings.
  • QA / Dev environments
    • CMPUTE can be used to launch your entire QA / Dev environments on Spot Instances as uptime constraints need not be to enforced strictly.
    • CMPUTE has introduced a new feature called “Scheduler” using which you can schedule the start time and end time of your environment. This assures you of your environments being up and running only for the specified duration, thereby ensuring you get the most savings.

Strategic Partnership with Cloudability

At AWS re:Invent 2016, we announced a strategic partnership with Cloudability, a financial management tool for monitoring and analyzing Cloud expenses. If you are an existing Cloudability customer, then you already know the power of precise recommendations that it provides to reduce your AWS costs. Using CMPUTE, you would now be able to act upon these recommendations in a frictionless manner. Using the tagging feature supported in Cloudability, you will be able to easily review and track cmpute performance. Extensive Cloudability reports and dashboards can be generated specific to cmpute jobs, clearly showing the increased usage of Spot Instances and the savings that it brings to the table.

So, there you go. You can now safely stick to your “reduce AWS cost by 85%” New Year’s resolution without breaking a sweat. Register now for a 14-day free trial of CMPUTE and see the results instantly.

Related Post

Leave a Reply

Your email address will not be published. Required fields are marked *

'