Fio benchmark

FioBenchmark

The fio benchmark job assists the user to directly run the fio command from the UI, avoiding the need to SSH into the client instance.

The job is preconfigured to run a read I/O benchmark against a single 20GB file that is assumed to be prepared as part of launching the FioLargeFileDemo cluster. Other benchmark parameters that are also included in the resulting fio command include:

• --direct=1

• --iodepth=1

• --ioengine=libaio

• --group_reporting

• --size=20G

• --readonly

Three options are offered as user configurable parameters:

• Block size: The size of each block chunk read or written

• Number of jobs: The number of concurrent jobs to run

• Read type: Either sequential read (read) or random read (randread); each are different I/O patterns.

The benchmark completion time can vary depending on the above parameters and if the data was previously cached in Alluxio or not (hot read vs cold read). After the benchmark is completed, the output will be shown as the job output.

Note the highlighted output information below the progress bar, noting the bandwidth of the operation and the duration. The raw output of the command is also shown.