Abstract:

Early researchers have observed that second level buffer caches commonly suffer from severe performance degradation, because the networked storage servers have special access patterns. To address the critical issue, this paper presents a novel cache replacement algorithm using a new mechanism called sequentiality detecting to achieve high performance. The attribute named sequentiality is defined for buffer pages to make a replacement decision: sequential pages with a high sequentiality value are preferred to be discarded rather than the random pages with a low sequentiality value. That is to say, this algorithm can reduce the numbers of random disk access and avoid unnecessary seeking and rotating cost of the disk heads. Our tracedriven simulation results show that it can significantly reduce effective access time and deliver a better performance for a wide range of cache sizes.

Key words: second level buffer cache;replacement policy;sequentiality detection