Issue No. 11 - November (2002 vol. 51)
<p><b>Abstract</b>—We address the problem of prefetching and caching in a parallel I/O system and present a new algorithm for parallel disk scheduling. Traditional buffer management algorithms that minimize the number of block misses are substantially suboptimal in a parallel I/O system where multiple I/Os can proceed simultaneously. We show that in the offline case, where a priori knowledge of all the requests is available, <scp>PC-OPT</scp> performs the minimum number of I/Os to service the given I/O requests. This is the first parallel I/O scheduling algorithm that is provably offline optimal in the parallel disk model. In the online case, we study the context of global <i>L</i>-block lookahead, which gives the buffer management algorithm a lookahead consisting of <i>L</i> distinct requests. We show that the competitive ratio of <scp>PC-OPT</scp>, with global <i>L</i>-block lookahead, is <tmath>\Theta (<i>M</i> - <i>L</i> + <i>D</i>)</tmath>, when <i>L</i> <u><</u><i>M</i>, and \Theta (<i>M</i> <i>D</i> / <i>L</i>), when <i>L</i> > <i>M</i>, where the number of disks is <i>D</i> and buffer size is <i>M</i>.</p>
Parallel I/O systems, caching, prefetching, scheduling, buffer management, competitive ratio, algorithms, online algorithm, offline algorithm.
M. Kallahalla and P. J. Varman, "PC-OPT: Optimal Offline Prefetching and Caching for Parallel I/O Systems," in IEEE Transactions on Computers, vol. 51, no. , pp. 1333-1344, 2002.