Shifting Performance Engineering Left
summary
Red Hat shifts performance testing into developers’ CI by embedding a Kubernetes-scale orchestration tool called Cubener/cubeer (create/delete/patch/read jobs, Prometheus scrapes, OpenSearch/Elastic indexing) so PRs and nightlies run over ephemeral clusters. The team runs 100+ weekly tests across five platforms (AWS, Azure, IBM cloud, bare metal, managed), exports metrics, and feeds them to Orion automatic regression (change‑point detection). This setup finds concrete regressions — e.g. a 900% kube API latency spike traced to a kernel change — and narrows faulty commits reliably.
AI-written summary · about
chapters (18)
- 0:00Introduction
- 1:26Late performance testing problems
- 3:56Goal of shifting left
- 4:58CI pipelines integration
- 6:03Ephemeral clusters strategy
- 7:25Platform achievements
- 7:46CubeBurner tooling
- 10:45CubeBurner config workflow
- 12:11OpenShift CubeBurner extensions
- 13:44CubeBurner job types
- 16:58CubeBurner alerting
- 17:39OpenShift workloads
- 19:28Post-processing Orion framework
- 21:13API latency regression case
- 23:21CubeBurner demo
- 29:38Grafana dashboard analysis
- 31:25Orion regression analysis
- 34:17Future steps
related talks
41mPerformance Engineering in the Age of DevOps and Agile
38mShift-Left Performance Engineering for RAG and LLM Platforms
18mOpenstack Continious Performance tests (CPT)
31mRelease Engineering as a Service - The Art of Integrating with K8s
33mPerformance Testing in Agile and DevOps
01h 08mPerformance testing in the cloud
33mMoving 75,000 Engineers To DevOps On The Public Cloud
44mContinuous Performance Testing in DevOps