APR: adaptive page replacement scheme for scientific applications

The amount of data in modern computing workloads is growing rapidly. Meanwhile, the capacity of main memory is growing slowly; thus, memory management of operating systems plays an increasingly important role in application performance. Recent scientific applications process large amounts of data as well. They tend to manage intermediate data in anonymous pages and repeat core operations on the data using loops. However, LRU variants have difficulty handling loop access patterns in scientific applications, which are commonly used as a page replacement policy in the operating system. In this article, we propose a new page replacement scheme, called adaptive page replacement (APR) for looping access patterns in scientific applications. APR can detect various looping access patterns and handle them appropriately online by exploiting the information already available in the virtual memory subsystem of OS. We evaluate APR by trace-driven simulation with traces extracted from 12 workloads in the SPLASH-2x benchmark. Throughout our experimental results, we demonstrate that APR outperforms prior schemes including CLOCK.

Automated summary

This article proposes a new page replacement scheme called Adaptive Page Replacement (APR) for looping access patterns in scientific applications. The experimental results show that APR outperforms prior schemes in terms of performance.