Proceedings of the 5th USENIX Conference on File and Storage Technologies (FAST '07), February 13–16, 2007, San Jose, CA. Supercedes Carnegie Mellon University Parallel Data Lab Technical Report CMU-PDL-06-106, May 2006.
Matthew Wachs, Michael Abd-El-Malek, Eno Thereska, Gregory R. Ganger
Dept. Electrical and Computer EngineeringServices that share a storage system should realize the same efficiency, within their share of time, as when they have the system to themselves. The Argon storage server explicitly manages its resources to bound the inefficiency arising from inter-service disk and cache interference in traditional systems. The goal is to provide each service with at least a configured fraction (e.g., 0.9) of the throughput it achieves when it has the storage server to itself, within its share of the server—a service allocated 1 nth of a server should get nearly 1 nth (or more) of the throughput it would get alone. Argon uses automaticallyconfigured prefetch/write-back sizes to insulate streaming efficiency from disk seeks introduced by competing workloads. It uses explicit disk time quanta to do the same for non-streaming workloads with internal locality. It partitions the cache among services, based on their observed access patterns, to insulate the hit rate each achieves from the access patterns of others. Experiments show that, combined, these mechanisms and Argon’s automatic configuration of each achieve the insulation goal.
FULL CONFERENCE PAPER: pdf
FULL TR: pdf
KEYWORDS: performance isolation, quality of service, shared storage, performance