dibble@cs.rochester.edu (08/19/89)
Contacts: Michael L. Scott scott@cs.rochester.edu Peter C. Dibble dibble@cs.rochester.edu Parallel computers with non-parallel file systems find themselves limited by the performance of the processor running the file system. We have designed and implemented a parallel file system called Bridge that eliminates this problem by spreading both data and file system computation over a large number of processors and disks. Our design will adapt easily to a wide variety of multiprocessors and multicomputers; our current implementation runs on the BBN Butterfly. To assess the effectiveness of Bridge we have used it as the basis for several standard file-handling applications, including copying, sorting, searching, and image transposition. Analysis and empirical measurements suggest that for applications such as these Bridge can provide nearly linear speedup on over 100 nodes.