MinFlow: High-Performance and Cost-Efficient Data Passing for I/O-Intensive Stateful Serverless Analytics

Explore a cutting-edge conference talk on MinFlow, a holistic data passing framework designed for I/O-intensive serverless analytics jobs. Delve into the challenges of serverless computing, particularly the "shuffle" operation in data analytics applications, and discover how MinFlow addresses performance degradation and high storage costs. Learn about the framework's innovative approach to generating multi-level data passing topologies, its interleaved partitioning strategy for optimizing function scheduling, and its precise model for determining optimal configurations. Gain insights into MinFlow's significant improvements over state-of-the-art systems like FaaSFlow and Lambada in terms of job completion time and storage cost. Presented by researchers from the University of Science and Technology of China and The Chinese University of Hong Kong, this 16-minute talk offers valuable knowledge for professionals and enthusiasts in serverless computing and data analytics.