Optimizing JVM for Large-Scale E-commerce Applications at Alibaba

Explore how Alibaba optimized the Java Virtual Machine (JVM) for their massive e-commerce applications running on over 100,000 servers. Dive into the customization and optimization of HotSpot (OpenJDK-based) to meet specific requirements for handling high-volume transactions. Learn about three key features implemented: deploying multiple containers in a single JVM instance, utilizing coroutines to reduce context switches, and implementing quick Java warmup. Gain insights into workload characterization, resource management, and the Loom project. Examine the Single API implementation, scheduling processes, and warmup techniques. Understand the considerations for engineer support and see a demonstration of the optimized JVM in action. This 44-minute talk by San Hong Li provides valuable knowledge for Java developers and system architects dealing with large-scale, high-performance applications.

Introduction
Agenda
Container Resource Management
Loom
Single API
Single API Demo
Selector
Schedule Process
Proposal
Implementation
Warmup
Application Context
Initializing
Results
Considerations
Engineer Support
Demonstration
Summary
QA