After several technical deep dives, the DoiT team identified the root causes of the problem: the current Build Release pipeline was causing delays and extended release cycles, leading to inefficiencies. As a result, DoiT recommended a multi-step approach. The plan included optimizing the Docker Container startup time and supporting the HackerRank team with code samples demonstrating Proof of Concept (PoC) and documentation. Additionally, DoiT planned on running unit tests in parallel, and moving the Build Server to a new setup in Elastic Container Services (ECS) with Fargate, more elastic and requiring lower maintenance.

The FinOps team at DoiT initiated an engagement to reduce the costs of AWS infrastructure. While HackerRank had already purchased Savings Plans (SPs), they they decided to leverage DoiT’s support and Spot Scaling technologies in addition to these SPs.

Using machine learning, DoiT’s Technology continuously monitors HackerRank’s cloud usage to identify additional compute instances not already covered. This increases the savings management that HackerRank works on without the need to purchase 1-year Savings Plans or Reserved Instances, or deal with the requisite forecasting and management challenges that would accompany them.

Additionally, Spot Scaling helped HackerRank identify an opportunity to optimize the costs of the Auto-Scaling Groups (ASGs) organising their EKS worker nodes in non-production environments. Managing the instance composition of their ASGs, Spot Scaling helped HackerRank adopt Spot instances and save on on-demand compute costs while dramatically reducing downtime risk.

Spot Scaling made best practice recommendations to their ASG configuration to allow for more instance types with similar attributes to their originally-selected instance type. In addition, Spot Scaling has a “fallback to on-demand” functionality, which automatically utilizes on-demand instances when there isn’t enough Spot capacity for their designated instance types. It enabled HackerRank to confidently modify many of their non-production ASGs to utilize Spot instances as much as possible.