Dear All, I am working on a micro-bechmarking code based on MPI. My program writes a 256 million integers array to a file using multiple processes. It reads the file back into the program using multiple processes too. The problem is that there is a huge difference between read and write performances.
Write Time: 182609 micro-seconds Read Time : 20196 micro-seconds The test system is a cluster comprising on 32 nodes and each node is based on intel xeon. The file system is Network File System(NFSv3). Has anyone else experience the same problem or can any one explain why there is this big gap b/w the read and write performance? Kind Regards -- Ammar ---------------------------------------------------------------------- Masters Student Dept. Of Computer Engineering, Kyung Hee University ----------------------------------------------------------------------
