// the find
hystax/optscale
FinOps and cloud cost optimization tool. Supports AWS, Azure, GCP, Alibaba Cloud and Kubernetes.
OptScale is a self-hosted FinOps platform that ingests billing data from AWS, Azure, GCP, Alibaba Cloud, and Kubernetes, then surfaces waste: idle VMs, unattached volumes, oversized instances, unused Reserved Instances. It's for platform/infra teams that want cloud cost visibility without paying for Apptio or CloudHealth, and have the ops maturity to run a Kubernetes cluster just to host their cost tool.
The recommendation engine is genuinely broad — rightsizing, RI/SP utilization, S3 lifecycle, Databricks cluster idle time, and power schedules for non-prod environments are all separate modules with their own archive/recommendation split, which suggests real operational history behind each one. The microservices architecture (bumiworker, diworker, trapper, herald, etc.) means individual data importers can fail or be updated without taking down the whole platform. CI coverage is thorough — there's a separate GitHub Actions workflow per service, so regressions in the AWS importer don't get masked by a passing auth test. The live demo is pre-seeded with realistic data, which is a low-friction way to evaluate before committing to the install.
The minimum hardware bar (8 cores, 16GB RAM, 150GB NVMe SSD) to run a cost optimization tool is a real irony — you're burning $50-100/month in infrastructure to find savings, which only pays off at meaningful cloud spend. The deployment process requires Ansible + Kubernetes + manually editing overlay YAML files, which is a significant ops burden for small teams who would benefit most from this. The database layer uses both MySQL (auth/alembic migrations) and MongoDB (resource data) and InfluxDB (metrics) and Redis — four data stores is a lot of surface area to operate and back up. Documentation lives entirely on an external commercial website (hystax.com), not in the repo, so if Hystax the company changes direction, the open-source install becomes unguided.