Explore a conference talk from OSDI '23 that introduces RON (Routing on Network-on-chip), a novel method for optimizing spinlocks in multi-core processors. Learn how this approach minimizes communication costs between cores using a routing table and pre-calculated locking-unlocking order. Discover how RON's one-way circular delivery of locks and data achieves both minimized global data movement cost and bounded waiting time. Examine the performance improvements RON offers in both user space and kernel space applications, including significant enhancements for Google LevelDB. Gain insights into how RON-plock addresses oversubscription issues with constant space complexity, outperforming existing solutions.
Overview
Syllabus
OSDI '23 - RON: One-Way Circular Shortest Routing to Achieve Efficient and Bounded-waiting Spinlocks
Taught by
USENIX