This course teaches learners how to implement Hydra, a resilient and highly available mechanism for remote memory. By utilizing erasure-coded remote memory with low latency, Hydra improves performance efficiency compared to existing solutions. The course covers the CodingSets algorithm for coding group placement and load balancing. Through Hydra, even with only 50% local memory, applications can maintain performance in the presence of remote failures. The teaching method includes discussions on the data plane, Hydra datapath for write and read operations, and achieving replication-like performance with lower overhead. This course is intended for individuals interested in enhancing remote memory resilience and performance optimization.
Overview
Syllabus
Intro
Remote Memory is Vulnerable
Remote Memory Performs Great!
Performance Tradeoff for Resiliency
Remote Memory w/ Hydra
Data Plane
Hydra Datapath - Asynchronously Encoded Write
Hydra Datapath – Late-binding Read
Replication-like Performance at Lower Overhead
Performance Tradeoff w/ Hydra
Taught by
USENIX
