jyvet/multirw

A multi-threaded program to stress a file system with parallel IOs in a single shared file.

MultiRW

This application spawns several threads to create multiple IO streams accessing a single file. It is designed to stress a file system with random parallel IOs to a single shared file.

How to build MultiRW

% make

How to run MultiRW

% ./multirw [OPTION...] <file>

• file: output file on the file system to stress

Options are :

-b, --bypass-cache=<value> Use O_DIRECT flag <0=disabled|1=enabled>
                           [default: 0]
-d, --duration=<seconds>   Time period the program should run [default: 10s]
-f, --file-size=<bytes>    File size [default: 512 MB + 1 byte]
-F, --multiple-fd=<value>  Open a file descriptor per thread
                           <0=disabled,1=enabled> [default: 0]
-i, --io-size-max=<bytes>  IO size max [default: 524288 bytes]
-l, --last-chunk=<bool>    Last IO should access last file chunk
                           <0=disabled,1=enabled> [default: 1]
-m, --mmap=<value>         Enable memory mapped file <0=disabled|1=enabled>
                           [default: 0]
-p, --pattern=<value>      IO pattern <0=read|1=write|2=rw> [default: 2]
-s, --seed=<value>         Initial seed [default: seed=PID]
-t, --threads=<value>      Amount of threads [default: 10]
-v, --verbose              Enable verbose mode
-?, --help                 Give this help list
    --usage                Give a short usage message
-V, --version              Print program version

Examples

Use 50 threads in write only mode:

% ./multirw -t 50 -p 1 /myfs/output_file.data

Use 100 threads, memory map the file and use O_DIRECT flag:

% ./multirw -t 100 -m 1 -b 1  /myfs/output_file.data

Use 20 threads and make the output file 1GB:

% ./multirw -t 20 -f $((1024**3)) /myfs/output_file.data