How Jenga & Tetris Can Help Run More Containers with Lesser Costs...
Lightningat ContainerCon @ OSS EU 2023 · on Tuesday, 2023-09-19· at 12:37· Room 3F (3JAUREGIA - Floor 3), Abandoibarra Etorb., 4, 48011 Bilbao, Bizkaia, Spain·
summary
Proposes a Kubernetes-focused rescheduler that simulates placement in memory and emits move-and-place recommendations using node/pod affinity APIs, with singular and multi-pod migration strategies that minimize total evicted capacity. Uses an entropy CPU/memory ratio to identify and swap imbalance-causing pods, accepts a scheduling time budget (example: 5 seconds) for pruning search, and reports a real test saving about $5,000/month (~30%) on clusters with >$2/hr nodes.
AI-written summary · about
chapters (7)
- 0:00Container scheduling challenges
- 1:25Tetris for resource fragmentation
- 3:31Impact of fragmented resources
- 4:57Rescheduling strategies
- 6:22Jenga for resource imbalance
- 8:55Cost savings from balance
- 10:18Runtime scheduling constraints
related talks
11mLightning Talk: How Jenga & Tetris Can Help Run More Containers with Lesser Costs & Downtime
58mHow Kubernetes plays Tetris with your containers
38mHow Kubernetes Plays Tetris with Your Containers
38mCost-Effective Scheduling of a Massive Number of Containers in Kubernetes
29mRight-Size Your Kubernetes Deployments
36mDon't Be Greedy: Rightsize Your Kubernetes Cluster With Prometheus
39mStaying in Tune: Optimize Kubernetes for Stability and Utilization
38mCost-Effective Scheduling of a Massive Number of Containers in Kubernetes