Resource Benchmarks

HPIO

The HPIO (High-performance I/O) benchmark is a tool for evaluating/debugging noncontiguous I/O performance for MPI-IO. It allows the user to specify a variety of noncontiguous I/O access patterns and verify the output. It has been optimized for PVFS2 MPI-IO hints, but can be augmented to use MPI-IO hints for other file systems.

Test cases:

Test case 1:

To test contiguous memory to contiguous file by using individual I/O.
Command: mpirun -np 4 ./hpio-small -n 1000 -m 10 -b 100

Test case 2:

To test contiguous memory to contiguous file, noncontiguous memory to contiguous file by using individual I/O.
Command: mpirun -np 4 ./hpio-small -n 1100 -m 10 -b 100

S3aSim

S3aSim, represents Sequence Similarity Search algorithm Simulator, is a sequence similarity search algorithm framework for testing and evaluating various I/O strategies using MPI-IO. It has been optimized for PVFS2 ROMIO hints, but can be augmented to use MPI-IO hints for other file systems. S3aSim uses a master-slave parallel programming model with database segmentation, which mimics the mpiBLAST access pattern.

Test cases:

Test case 1a: test serial I/O

To test individual I/O.
Command: mpirun -n 4 ./s3asim-nj -i 0

Test case 1b: test serial I/O

To test collective I/O.
Command: mpirun -n 4 ./s3asim-nj -i 1

Test case 2: test parallel I/O

To test parallel I/O.
Command: mpirun -n 4 ./s3asim-nj -p 1

Noncontig Benchmark

This is the benchmark that generates a regular, strided, non-overlapping access pattern perfectly suited for collective buffering.

Test case:

To test access pattern perfectly suited for collective buffering.
Command: mpirun -n 2 ./noncontig -fname /ufs3/home/scs/lfeng/benchmarks/s3asim-1.00/s3asim-test_1