Date of Original Version
All Rights Reserved
Abstract or Description
Modern high-end disk arrays typically have several gigabytes of cache RAM. Unfortunately, most array caches employ management policies in which the same data blocks are cached at both the client and array levels of the cache hierarchy - that is, they are inclusive. As a result, the aggregate cache behaves as if it was only as big as the larger of the client and array caches, instead of as large as the sum of the two.
This paper explores the potential benefits of exclusive caches, in which data blocks are either in the client or array cache, but never in both. Exclusivity helps to create the effect of a single, large unified cache. We propose an operation called 'demote' for transferring data ejected from the client cache to the array cache, and explore its effectiveness in increasing cache exclusivity using simulation studies. We quantify the benefits of demote, the overhead it adds, and the effects of combining it with different cache replacement policies across a variety of workloads. The results show that we can obtain useful speedups for both synthetic and real-life workloads.