SONIC - Application-Aware Data Passing for Chained Serverless Applications
Offered By: USENIX via YouTube
Course Description
Overview
Explore a 16-minute conference talk from USENIX ATC '21 that delves into SONIC, an innovative application-aware data passing system for chained serverless applications. Learn about the challenges of data exchange in serverless environments and how SONIC optimizes performance and cost by dynamically selecting the most efficient data-passing method. Discover the comparison between VM-Storage, Direct-Passing, and Remote-Storage approaches, and understand how factors like input data size, intermediate data size, application parallelism, and network bandwidth influence the optimal choice. Gain insights into SONIC's integration with Open-Lambda and its performance evaluation on Amazon EC2 using three popular analytics applications. Examine the system's ability to provide lower latency and higher performance per dollar across diverse conditions compared to baselines like SAND, vanilla OpenLambda, OpenLambda with Pocket, and AWS Lambda.
Syllabus
Intro
Background: Serverless Computing DAG & Data Passing
Serverless Data-Passing Challenges
SONIC's Data-Passing Alternatives
Data-passing performance trade-off
Direct-Passing vs Remote Storage With higher network bandwidth, the crossover point between Direct passing and Remote-passing shifts to higher fanout values
OUR SOLUTION: SONIC Hybrid Data-Passing Approach
Greedy Data-Passing Decisions: Pitfalls
SONIC: Design Overview
Evaluation: Baselines
Evaluation: E2E Latency and Cost
Content Sensitivity (Cont.)
Evaluation: Content Sensitivity
Evaluation: Scalability
Conclusion
Taught by
USENIX
Related Courses
Understanding China, 1700-2000: A Data Analytic Approach, Part 1The Hong Kong University of Science and Technology via Coursera The Analytics Edge
Massachusetts Institute of Technology via edX 大数据与信息传播 Big Data and Information Dissemination
Fudan University via Coursera The Future of Fashion
Marist College via Independent The Mobile Consumer
Marist College via Independent