Integrating Theory and Practice in Parallel File Systems

Several algorithms for parallel disk systems have appeared in the literature recently and they are asymptotically optimal in terms of the number of disk accesses Scalable systems with parallel disks must be able to run these algorithms We present for the rst time a list of capabilities that must be provided by the system to support these optimal algorithms control over declustering querying about the con guration independent I O and turning o parity le caching and prefetching We summarize recent theoretical and empirical work that justi es the need for these capabilities In addition we sketch an organization for a parallel le interface with low level primitives and higher level operations