Data Sieving and Collective I/O in ROMIO
FRONTIERS '99 Proceedings of the The 7th Symposium on the Frontiers of Massively Parallel Computation
IPDPS '09 Proceedings of the 2009 IEEE International Symposium on Parallel&Distributed Processing
Hi-index | 0.00 |
This paper presents a framework, Orthrus, that can accommodate multiple collective-I/O implementations, each optimized for some performance aspects, and dynamically select the best performing one accordingly to current workload and system performance bottleneck. We have implemented Orthrus in the ROMIO library. Our experimental results with representative MPI-IO benchmarks show that Orthrus can significantly improve the performance of collective I/O under various workloads and system scenarios.