IJCSNS International Journal of Computer Science and Network Security, v.7, no.3, pp.155 - 160
Abstract
Performance studies of buffer caches have generally been done with write references ignored. In this study, we show that once we consider write references in buffer caches that include nonvolatile memory, the traditional hit ratio performance measure is no longer an accurate representation, and that the disk access ratio should be used. We also show that in regards to this measure, the MIN replacement algorithm is either non-optimal or non-applicable.
Publisher
International journal of computer science and network security