Title
An Implementation and Evaluation of Client-Side File Caching for MPI-IO
Abstract
Client-sidefile caching has long been recognized as afile system enhancement to reduce the amount of data transfer between application processes and I/O servers. However, caching also introduces cache coherence problems when a file is simultaneously accessed by multiple processes. Ex- isting coherence controls tend to treat the client processes independently and ignore the aggregate I/O access pattern. This causes a serious performance degradation for paral- lel I/O applications. In our earlier work, we proposed a caching system that enables cooperation among applica- tion processes in performing client-side file caching. The caching system has since been integrated into the MPI-IO library. In this paper we discuss our new implementation and present an extended performance evaluation on GPFS and Lustre parallel file systems. In addition to compar- ing our methods to traditional approaches, we examine the performance of MPI-IO caching under direct I/O mode to bypass the underlying file system cache. We also investi- gate the performance impact of two file domain partition- ing methods to MPI collective I/O operations: one which creates a balanced workload and the other which aligns accesses to the file system stripe size. In our experiments, alignment results in better performance by reducingfile lock contention. When the cache page size is set to a multiple of the stripe size, MPI-IO caching inherits the same advantage and produces significantly improved I/O bandwidth.
Year
DOI
Venue
2007
10.1109/IPDPS.2007.370239
IPDPS
Keywords
Field
DocType
message passing,application software,data structures,file servers,message passing interface,memory management,coherent control,cache coherence,data transfer,computer science,control systems
Distributed File System,File system,Self-certifying File System,Computer science,Cache,Parallel computing,Device file,Torrent file,Memory-mapped file,File system fragmentation,Operating system,Distributed computing
Conference
Citations 
PageRank 
References 
30
1.68
15
Authors
5
Name
Order
Citations
PageRank
Wei-keng Liao1109587.98
Avery Ching222116.21
Kenin Coloma31379.36
Alok N. Choudhary424222.44
Lee Ward51859.81