StorageOS compression is handled on a per volume basis and is enabled by
default, as performance is generally increased when compression is enabled due
to fewer read/write operations taking place on the backend store (the volumes’
blob files). Compression can be disabled
by setting the label
on a volume.
StorageOS utilises the lz4 compression algorithm
when writing to the backend store and when compressing replication
traffic before it is sent across the network.
Compression is granular per 4k block and data will remain
compressed/uncompressed once written to a volume. Therefore, compression can be
dynamically enabled and disabled by setting the
label on a volume.
StorageOS detects whether a block can be compressed or not by creating a heuristic that predicts the size of a compressed block. If the heuristic indicates that the compressed block is likely to be larger than the original block then the uncompressed block is stored. Block size increases post compression if the compression dictionary is added to a block that cannot be compressed. By verifying whether blocks can be compressed, disk efficiency is increased and CPU resources are not wasted on attempts to compress uncompressible blocks. StorageOS’ patented on disk format is used to tell whether individual blocks are compressed without overhead. As such volume compression can be dynamically enabled/disabled even while a volume is in use.
When compression and encryption are both enabled for a volume, blocks are compressed then encrypted.