proteanTecs Slashes K8s Costs with AI-Driven Recommendations

proteanTecs deployed PerfectScale by DoiT, which compares allocated resources against actual utilization for each service to identify over-provisioning and recommend safe, specific actions to cut costs. The team first rolled it out in a development environment, then expanded into production—using PerfectScale's adjustable 'resilience level' to give critical services additional headroom and ensure stability. PerfectScale also surfaced under-provisioned services suffering from throttling, OOM kills, and container restarts that had gone undetected by existing monitoring tools.

proteanTecs is on a mission to enable the electronics industry to scale. Its cloud-based platform combines deep telemetry data with machine learning to monitor the health and performance of advanced chips, transforming the way the reliability of electronics is achieved. proteanTecs solutions are a critical component used by key customers across the semiconductor, data center, and automotive industries—requiring an application infrastructure that is stable, scalable, and highly available. To meet these demands, proteanTecs runs its entire platform on Kubernetes, with more than 30 clusters spanning AWS, Azure, and Google Cloud.

As proteanTecs grew in both functionality and end-user adoption, the early focus on stability and availability gave way to a new concern: sub-optimized resources that were inflating the cloud bill. 'Our development team builds their microservices and chooses their own resource request and limits,' explained Amit Daniel, Director of DevOps at proteanTecs. 'However, the developers tended to allocate risk-averse resources to ensure their services were working as expected. This approach kept our systems available and stable, but it was inefficient from a cost standpoint.' The DevOps team was tasked with optimizing the K8s environment, but their existing tools had significant gaps. 'With tools like DataDog, we can see whatever we want; requests, limits, etc., but they only give us visibility. We needed better insights into what resources our microservices need to run properly,' Daniel said.

After being introduced to PerfectScale by DoiT, the proteanTecs team deployed it in a development environment and immediately began identifying optimization opportunities. 'Our Dev environment was costing roughly $10k per month,' said Daniel. 'By implementing PerfectScale's recommendations, we were able to get the cost down to $5k without compromising the stability of the environment.' As the team rolled the solution into production, they used PerfectScale's adjustable 'resilience level' to grant critical services additional headroom—ensuring cost optimization efforts would not impact system performance.

Beyond cost savings, PerfectScale surfaced reliability issues that had been flying under the radar. 'We found throttling, OOM, and container restarts that we did not know were happening. Many issues were not triggering alerts in our monitoring system. Also, if we were not capturing the right metrics, it was challenging and time-consuming to determine the root cause of something like a memory leak,' Daniel explained. 'PerfectScale instantly identifies the issues and shows us how to resolve them quickly. As a result, we see that our platform is performing better, which is helping us to provide our customers with a better experience.'

The DevOps team is currently leading optimization efforts, but plans to extend ownership to developers. 'I have PerfectScale opened on my browser at all times, but after we do the initial phase of cost optimization, we plan on expanding the scope to our developers,' Daniel said. The goal is to embed cost awareness and accountability across engineering, with developers expected to evaluate the cost-effectiveness of their services after every release and deployment. 'As we continue to deploy new innovations to our application, and as the customer base grows, we want to make sure our environment grows as efficiently as possible,' Daniel said. 'PerfectScale will be an essential solution in helping us keep our environment cost-effective and stable.'