A K8s-Based Workload Allocation Optimizer for Minimizing Power Consumption

on YouTube ↗24m

" summary

Describes WAO, a Kubernetes-based Workload Allocation Optimizer that uses machine learning to predict workload-driven power increases and adds a scoring plugin to the Kubernetes scheduler for node selection. WAO includes a load balancer component that assigns Pods to Nodes aiming to minimize data center power use while controlling service performance, and the system is demonstrated on a real edge data center of 200+ servers rather than in a virtual environment.

AI-written summary · about

" chapters (23)

1. 0:00Intro
2. 1:34Who are we: Matsuoka Lab at Osaka Univ.
3. 2:37Our technology: Liquid immersion cooling
4. 2:51Our technology: optimal task allocation
5. 3:56The complexity of edge computing systems
6. 4:20Spread of Microservices & Power consumption increases
7. 6:34Our Approach: WAO on K8s
8. 8:01WAO power saving operation strategy
9. 8:38Scheduler Framework on K8s
10. 9:55Architecture of WAO Based Scheduler
11. 10:51Architecture of WAO Based Load Balancer
12. 11:35Setting of Machine Learning Power consumption (PC) model
13. 12:41How does it work? How is the performance? Environment
14. 13:45Preset Temperature of Air Conditioner
15. 15:02Power Consumption Model
16. 15:42Evaluation Value in WAO Based Load Balancer
17. 16:56Evaluation of Power Consumption Reduction
18. 17:35Evaluation of "WAO Scheduler" + MetalLB
19. 19:03Evaluation of Kube Scheduler + "WAOLB"
20. 20:28Evaluation of complete K8s-WAO solution
21. 21:10Evaluation of Kubernetes-based WAO
22. 21:51Evaluation of Response Time
23. 22:55Conclusion

" related talks

• 11mK8s Resources Ensuring Performance and OptimizationPedro Alpis · KCD Rio de Janeiro 2025
• 29mRight-Size Your Kubernetes DeploymentsRémi Verchere · Devoxx France 2024
• 59mHow Kubernetes optimisation can combat climate changeAnnie Talvasto · NDC London 2023
• 28mIntelligent Workload Re-shuffling to Minimize Idle Resources Across ClusterYash Bhatnagar, Hemani Katyal · Kubernetes Forum Bengaluru 2020
• 07mLightning Talk: Environmentally Sustainable AI via Power-Aware Batch SchedulingAtanas Atanasov · Kubernetes AI + HPC Day North America 2023
• 05mResource Roulette: Winning the Kubernetes Allocation GameDaniele Polencic · KubeCon + CloudNativeCon Europe 2025
• 34mPredictive autoscaling patterns in KubernetesRoberto Carratala, Germán M Yébenes · ContainerDays 2022
• 35mSustainability Research the Cloud Native WayChen Wang, Huamin Chen · KubeCon + CloudNativeCon North America 2022