Energy-Aware Workload Orchestration on OCP Servers using Kubernetes.pdf

《Energy-Aware Workload Orchestration on OCP Servers using Kubernetes.pdf》由会员分享，可在线阅读，更多相关《Energy-Aware Workload Orchestration on OCP Servers using Kubernetes.pdf（23页珍藏版）》请在三个皮匠报告上搜索。

1、ECO-QubeEAWA:energy-awareworkloadassignment(EAWA)algorithmfortheminimization of IT energy consumption at datacentersEnergy-Aware Workload Orchestration on OCP Servers using KubernetesaatayYlmaz,Project Manager,Research Institutes of RISEDr.Ender Demirel,CEO,Design&Simulation Tech.Ouzhan Herkilolu,Sr

2、.DevopsEngineerEnergy-Aware Workload Orchestration on OCP Servers Using KubernetesDC SUSTAINABILITYSUSTAINABILITY State of the art of existing cooling systems Power model for OCP Servers Fast thermal evaluation in a data center Energy-Aware Workload Assignment Implementation in Kubernetes Cooling op

3、timization&energy minimization in data centers EU regulations for energy-aware data centers OutlineAre we over-cooling?During the Last Glacial Maximum(LGM),which occurred approximately 21,000 years ago during the last Ice Age,average global temperatures were about 4 to 7 degrees Celsius(7 to 13 degr

4、ees Fahrenheit)coolerthan today.Do you think a 4-degree difference is significant?Are we excessively cooling?Source:NatureState of the Art of Existing Cooling SystemsCompute elementEnvironmentState of the Art of Existing Cooling SystemsSource:Raspberry PiSource:ASHRAEState of the Art of Existing Coo

5、ling SystemsTools Developed in ECO-QubeEstimatesserverpowerconsumptionconsideringCPUutilization,CPUtemperature and fan speed.Distributesworkloadsacrossservers in a way that minimizesIT energy consumption.Predicts outlet and CPU temperatures of the servers for a given workload distribution.Optimizes

6、cooling system fortheminimizationofthecooling energy.ECO-SERVERPOWER:A Power Model for OCP ServersServer Model:Leopard V2 OCPFan Model:Delta PFR0812DHE-F00Maximum RPM:9200Power Consumption:25.2 W2 CPUs56 threadsHyper-threading is active Variations of server power with the CPU utilization for various