Academic
Publications
Noncontiguous I/O through PVFS

Noncontiguous I/O through PVFS,10.1109/CLUSTR.2002.1137773,Avery Ching,Alok N. Choudhary,Wei-keng Liao,Robert B. Ross,William Gropp

Noncontiguous I/O through PVFS   (Citations: 48)
BibTex | RIS | RefWorks Download
With the tremendous advances in processor and memory technology, I/O has risen to become the bottle- neck in high-performance computing for many applica- tions. The development of parallel file systems has helped to ease the performance gap, but I/O still remains an area needing significant performance improvement. Research has found that noncontiguous I/O access patterns in scien- tific applications combined with current file system meth- ods to perform these accesses lead to unacceptable per- formance for large data sets. To enhance performance of noncontiguous I/O, we have created list I/O, a native ver- sion of noncontiguous I/O. We have used the Parallel Virtual File System (PVFS) to implement our ideas. Our research and experimentation shows that list I/O outper- forms current noncontiguous I/O access methods in most I/O situations and can substantially enhance the perform- ance of real-world scientific applications.
Conference: Cluster Computing - CLUSTER , vol. cs.DC/0207, pp. 405-414, 2002
Cumulative Annual
View Publication
The following links allow you to view full publications. These links are maintained by other sources not affiliated with Microsoft Academic Search.
    • ...In order to closely match the List I/O interface proposal introduced in [20], we introduce a series of versioning-oriented primitives that facilitate non-contiguous manipulations of data at the level of BLOBs...

    Viet-Trung Tranet al. Efficient support for MPI-I/O atomicity based on versioning

    • ...Existing studies on noncontiguous I/O optimizations [18], [19], [20], [21], [22], [23], on the other hand, emphasize improving actual I/O performance in isolation and often ignored interactions with the scheduling of computation...
    • ...In particular, PVFS (on IA64) provides special optimization for this write pattern with LIST I/O technique [21]...

    Heshan Linet al. Coordinating Computation and I/O in Massively Parallel Sequence Search

    • ...These primitives are closely matched the List I/O interface proposal introduced in [14]...

    Viet-Trung Tran. Towards a Storage Backend Optimized for Atomic MPI-I/O for Parallel Sc...

    • ...Current solutions to the noncontiguous I/O of parallel file system for “cluster” platform involve calling multiple independent I/O requests, using “data sieving” I/O [7] or using list I/O [8]...
    • ...The current strategies used to handle these noncontiguous situation involve call multiple independent I/O requests, using “data sieving” techniques to take advantage of the high transfer rate of larger disk operations [7] and PVFS’s “list I/O” [8] that would reduce I/O accesses, independent of the...
    • ...PVFS enhances its I/O request with the description of a “list”, which represents the noncontiguous memory blocks and disk blocks [8]...
    • ...The other patterns, such as block-block pattern [8], might produce different results, but the critical factor will still be the data disperse level...

    Qiuming Luoet al. Performance analysis for noncontiguous I/O using multiple I/O, data si...

    • ...PVFS is an open source parallel file system designed for computational science applications [17, 18, 19, 20]...

    Samuel Langet al. I/O performance challenges at leadership scale

Sort by: