Parallel I/o Scheduling and Buffer Management

Parallel I/O systems are an integral component of modern high performance systems, providing large secondary storage capacity, and having the potential to alleviate the I/O bottleneck of data intensive applications. In these systems the I/O buffer can be used for two purposes (a) improve I/O parallelism by buffering prefetched blocks and making the load on disks more uniform, and (b) improve I/O latency by caching blocks to avoid repeated disk accesses for the same block. To make best use of available parallelism and locality in I/O accesses, it is necessary to design and implement prefetching and buffer management algorithms that schedule reads intelligently so that the most useful blocks are prefetched into the buffer and the most valuable blocks are retained in the buffer when the need for evictions arises. This dissertation focuses on prefetching and buffer management algorithms for parallel I/O systems. Our aim is to exploit the high parallelism provided by multiple disks by using appropriate buffer management to reduce the average read latency seen by an application. The thesis is that, prefetching and buffer management in parallel I/O systems is fundamentally different from that in systems with a single disk, thereby necessitating new algorithms to handle it. To this end we first present evidence to show the limitations of intuitive algorithms that generalize those for sequential systems to a parallel I/O system, and then design algorithms that have better performance.