Explore techniques for optimizing MySQL container resources in Kubernetes environments through this conference talk. Learn how JD.com developed a system combining statistical analysis, forecasting, and optimization algorithms to dynamically adjust container resource allocations and reschedule containers via Kubernetes and Vitess APIs. Discover methods for workload characterization, right-sizing CPU resources, and implementing auto-scaling strategies. Examine multi-resource balance approaches for host selection and correlation-aware techniques. Gain insights into experimental evaluation setups and key results that demonstrate significant improvements in resource efficiency and cost reduction for large-scale MySQL deployments supporting e-commerce services.
Right-Sizing and Auto-Scaling of MySQL Containers in Kubernetes
Overview
Syllabus
Intro
JD Elastic Database
Problems and Challenges
Workload Characterization
Right Sizing Estimate the worldoad demand
Right Sizing of CPU Resources
Auto Scaling and Rescheduling
Auto-Scaling: Overview and Cost Models
Host Selection: Multi-Resource Balance
Experimental Evaluation Setup
Host Selection: Resource Availability
Host Selection: Correlation-awareness
Key Results
Conclusions
Taught by
Linux Foundation
