Batch-Size Optimization in Velox Aggregation: Performance Analysis and Memory Management

Explore a 13-minute technical presentation from Bytedance that delves into HashAggregation optimizations discovered through investigating batch-size impacts in Velox aggregation. Learn how an unexpected finding about batch-size differences led to valuable insights into memory management efficiency and performance optimization techniques. Gain practical knowledge about analyzing performance variations, forming hypotheses, and validating theories through experimental methods. Discover the ongoing developments in Velox, the Native Presto Worker, and understand the engineering efforts driving continuous improvements in the system. Perfect for developers and engineers interested in database optimization, particularly those working with Presto and Velox technologies.