# fs-benchmark **Repository Path**: hackydh/fs-benchmark ## Basic Information - **Project Name**: fs-benchmark - **Description**: perftest fs - **Primary Language**: Python - **License**: MIT - **Default Branch**: master - **Homepage**: None - **GVP Project**: No ## Statistics - **Stars**: 0 - **Forks**: 0 - **Created**: 2019-05-20 - **Last Updated**: 2020-12-19 ## Categories & Tags **Categories**: Uncategorized **Tags**: None ## README # fs-benchmark This benchmark supports read-write and metadata performance test on the wide distribute file system, such as [ceph](https://ceph.com), [moosefs](https://moosefs.com), [lizardfs](https://lizardfs.org/) and so on and generates test graphs using pandas. # read-write test This test uses [iozone](http://iozone.org/) to run the read-write test and includes write, read, randread, randwrite. ## default parameters ``` block size(KiB): 8 16 32 64 128 256 512 1024 2048 concurrent: 1 2 4 8 16 file size: 1GiB ``` ## run and plot ```bash cd /path/to/mount/fs NAME=test TEST_LOOP=3 DIRECT="" bash $ROOT/rw-benchmark/iozone/perftest.sh | tee /tmp/testlog # the result files is in $HOME dir python $ROOT/rw-benchmark/plot/plot.py -f $HOME/iozone-test--{}.tar.gz -r 3 ``` # metadata test metadata test tests the performance of MDS of ceph, or master of moosefs, or other fs ``` Usage of perftest: -b uint start index of operations -c uint concurrent of operations (default 1) -id uint client id (default 1) -interval duration interval to calculate tps, e.g. 1h5m8s (default 1m0s) -n string count string of operations, k,m,g (default "10") -period duration period duration of each stages, e.g. 1h5m8s (default 5m0s) -s string count of operations in one dir, k,m,g (default "10") -t for test -v show version ``` ## Example ``` cd /path/to/mount/fs $ROOT/meta-benchmark/bin/perftest -n 3m -c 100 -id 1 -interval 1m -period 3h -n 3,000,000 files -c 100 concurrent -id 1 -interval 1 minute -period 3 hour ```