Proceedings of the 10th USENIX Symposium on Networked Systems Design and Implementation (NSDI'13), Lombard, IL, April 2013. Supersedes Carnegie Mellon University Parallel Data Lab Technical Report CMU-PDL-12-116. November 2012.
Bin Fan, David G. Andersen, Michael Kaminsky*
School of Computer Science
Carnegie Mellon University
*Intel Labs
This paper presents a set of architecturally and workloadinspired algorithmic and engineering improvements to the popular Memcached system that substantially improve both its memory eciency and throughput. These techniques—optimistic cuckoo hashing, a compact LRU-approximating eviction algorithm based upon CLOCK, and comprehensive implementation of optimistic locking—enable the resulting system to use 30% less memory for small key-value pairs, and serve up to 3x as many queries per second over the network. We have implemented these modifications in a system we call MemC3—Memcached with CLOCK and Concurrent Cuckoo hashing—but believe that they also apply more generally to many of today's read-intensive, highly concurrent networked storage and caching systems.
KEYWORDS: hashing, performance, memory efficiency, key-value store
FULL TR: pdf