Environmental Sustainability: An Energy-Aware Datacenter Operating System
Talkat KubeDay Japan 2022 · on Wednesday, 2022-12-07· at 11:40· Level 3 | InterContinental Ballroom·
summary
A data center OS stacks Zabbix and Prometheus for 4,000 signals per minute, feeds Apache Kafka and ksqlDB for realtime cleaning, trains per-device machine learning models (server and AC) with MFO-style experiments, and extends the Kubernetes scheduler to score Pod-to-Node placement; tested on a 200+ server testbed running an object‑detection workload and achieves about 13% power reduction while balancing latency. Key components: Zabbix and Prometheus, Kafka stream processing, Kubernetes scheduler extension, 200 servers.
AI-written summary · about
chapters (7)
- 0:00Start
- 0:44Outline
- 4:18Background
- 10:59Research Data collection
- 14:39Power consumption of Data center
- 15:49Power consumption model
- 17:17How to build?
related talks
35mSustainability Research the Cloud Native Way
30mSustainability Chronicles: Innovate Through Green Technology
29mEdge DC Energy Efficiency: A K8s Workload Allocation Optimizer with Generic Server Power Model
39mHow we learned to love the Data Center Operating System
27mGreening the AI Cloud: Validating Power Models for Kubernetes Containers
40mSustainability the Container Native Way
19mKeynote: Operating 600+ Mesos Servers on 7 Datacenters @Criteo
14mOpen Source Ecosystem: Building Sustainable Solutions with Kubernetes