The Community for Technology Leaders
Parallel Architectures, Algorithms and Programming, International Symposium on (2011)
Tianjin, China
Dec. 9, 2011 to Dec. 11, 2011
ISBN: 978-0-7695-4575-2
pp: 218-222
For parallel disk array systems, the parallelism among disks is the key factor influencing the performance and the scale of systems. Unfortunately, the parallelism of cached blocks is largely ignored by cache management schemes that focus on reducing the number of cache misses. Therefore, the performance of parallel disks array systems for workloads with a skew access pattern can be seriously degraded. To solve this problem, we propose a Parallelism based Cache Replacement scheme (PCAR) for parallel disks array systems, which can exploit both of the inter-disks parallelism and the intra-disk spatial locality. We have implemented the prototype of PCAR algorithm in Linux 2.6.18. And, the experimental results show that PCAR outperforms DULO and LRU by up to 22.8% and 33.1% in terms of the average response time, and by up to 20% and 43.9% in terms of throughput.
Xiaodong Shi, Dan Feng, "PCAR: Parallelism Based Cache Replacement Scheme to Exploit Inter-disks Parallelism and Intra-disk Spatial Locality in Parallel Disk Array", Parallel Architectures, Algorithms and Programming, International Symposium on, vol. 00, no. , pp. 218-222, 2011, doi:10.1109/PAAP.2011.69
91 ms
(Ver 3.3 (11022016))