Smart Green Computing Cloud Native Operations
summary
Proof-of-concept runs on a three-node cluster whose nodes simulate power from coal, petroleum and natural gas, enforcing a per-node CO2 threshold (demo target ~220 units, examples use 100-unit policy checks). Metrics flow from Kepler, Telegraf and collectd into Prometheus, an XGBoost model in OpenDataHub emits “Smart Ops” metrics via a Python metrics proxy, and a secondary Kubernetes scheduler with Intel task policies evicts and reschedules pods to keep CO2 under the goal — the demo successfully migrates pods off violating nodes.
AI-written summary · about
chapters (11)
- 0:00Intro
- 1:19Goal-driven Workload Management (1)
- 3:18Domain-aware Workload Management
- 4:44Adopt GitOps Model
- 6:29Adopt Enhanced Visibility
- 7:49Adopt Intent Driven Orchestration
- 9:52Example of Use Cases for Stack
- 14:15PoC High-Level Architecture
- 15:50Metrics Pipeline
- 19:09Governance/Scheduling
- 28:22Referenced Projects
related talks
21mSustainable Cloud Operations
12mKeynote: Cloud Native Sustainability - How You Can Help Right Now!
28mMore Data Please: Hands on Green Cloud Experiments
20mMonitoring Energy Consumption and Building Energy Efficient Systems the Cloud Native Way
28mSustainability and Cloud Native: Strategies for a Green Future
20mEnergy Efficient Placement of Edge Workloads
42mUndertaking a GreenOps Approach in the Public Cloud
59mHow Kubernetes optimisation can combat climate change