Creating a filesystem in AWS ParallelCluster is easy, we can directly specify the filesystem type, options and mount location. This gives several advantages over creating the filesystem outside of AWS ParallelCluster:
In the lab, we setup the filesystem with the following options:
FSx Lustre Parameter | Definition |
---|---|
DataCompressionType | Data Compression both lowers storage size and increases throughput, set this to LZ4 |
DeploymentType | Options are SCRATCH_1 , SCRATCH_2 , PERSISTENT_1 , and PERSISTENT_2 . I reccomend using persistent as opposed to scratch when mounting through AWS ParallelCluster. Scratch is ideal for per-job filesystems. |
PerUnitStorageThroughput | Options are 125 , 250 , 500 or 1000 MB/s/TiB of throughput. |