SAN-level caching can manage caching within a global view so that global hot data can be identified and cached. However, two problems may be encountered in the existing SAN-level caching methods: first, too many migrations from disks to cache resulting from cache thrashing; second, too few cache hits resulting from insufficient caching. This paper proposes a SAN-level caching method employing chunk-aging to control the migrations of chunks of data from disks to cache, which takes the temporal distribution of chunk accesses into account and solves the two problems above. Then by employing two LRU lists to manage the cache replacement, this method separates and manages two types of hot data chunks, one of which have burst accesses and the other have long-term frequent accesses, thus preventing them from interfering with each other in cache. The simulation results demonstrate that our method can achieve a high hit ratio while maintaining a relatively low migration rate.
展开▼
