Ph.D. Thesis Proposal
Recent studies on operating system support for concurrent server systems mostly target CPU-intensive workloads with light disk I/O activities. However, an important class of server systems that access a large amount of disk-resident data, such as the index searching server of large-scale Web search engines, has received limited attention. In this thesis work, we plan to examine operating system techniques to improve the performance of data-intensive server systems under concurrent execution. We will propose OS enhancements in three aspects of the operating system: file system prefetching, memory management, and disk I/O management. First, we will propose a competitive prefetching strategy that can balance the overhead of disk I/O switching and the wasted I/O bandwidth of prefetching unnecessary data. Second, we plan to explore a new memory management method for prefetched data, in order to reduce prefetching-incurred page thrashing at high concurrency levels. Third, we plan to investigate an I/O cancellation strategy that can improve the performance of data-intensive server systems by partially canceling unnecessary disk I/O prefetching.
We have partially implemented the first two proposed techniques in the Linux 2.6 kernel. Preliminary evaluation on microbenchmarks and real applications shows promising results. Based on the preliminary work, we will provide future research plan for our proposed OS enhancements.