Using reference counters in update based coherent memory

Abstract

As the disparity between processor and memory speed continues to widen, the exploitation of locality of reference in shared-memory multiprocessors becomes an increasingly important problem in parallel processing. In this paper, we explore the problem of managing locality at the operating system level. In specific, we study the use of reference counters in making informed decisions about page placement and movement. We use trace-driven simulation of real applications to evaluate the effectiveness of reference counters in providing useful hints to the memory manager of the operating system. Our main conclusion is that reference counters provide a simple and inexpensive mechanism for detecting the reference patterns of pages and making robust page placement decisions that result in significant performance improvement.