Guessing a workload's memory limit takes thirty seconds. Living with that guess takes months.
@picnic_ai helps pharmaceutical researchers run observational clinical trials, processing large volumes of medical data across several Kubernetes clusters. CPU and memory needs shift job to job, but resource requests were set at creation and rarely touched again until the out-of-memory errors arrived. Jobs failed and retried, queues built up, shared databases took the load.
PicnicAI came to
@PerfectScale_io by DoiT through a broader cloud cost optimization initiative. Our Field Engineering and Customer Success teams worked with their engineers on how HPA and the cluster autoscaler interact, and which workloads were safe to automate first.
PerfectScale by
@doitint carried the optimization: PerfectScale's InfraFit exposed node configurations that did not fit their workloads, and automation policies let them widen rightsizing at their own pace across development, staging and production.
✅ 85-90% of application workloads on automated resizing, up from zero
✅ ~50% less team time spent managing infrastructure
✅ Fewer memory-related failures across variable data-processing workloads
Read the full story:
bit.ly/4izrE5Y