Difference between revisions of "Performance tuning"

Jump to navigation Jump to search
m
m (Added redundant_metadata=mostly recommendation for steam games.)
Line 195: Line 195:
== Sequential workloads ==
== Sequential workloads ==


Set recordsize=1M on datasets that are subject to sequential workloads. Note that the large_blocks feature must be enabled on the pool to use larger record sizes than 128K (the default). `zfs send` operations must specify -L to ensure that larger than 128KB blocks are sent and the receiving pools must support the large_blocks feature.
Set recordsize=1M on datasets that are subject to sequential workloads. Note that the large_blocks feature must be enabled on the pool to use larger record sizes than 128K (the default). The large_blocks feature is enabled by default on new pools. `zfs send` operations must specify -L to ensure that larger than 128KB blocks are sent and the receiving pools must support the large_blocks feature.


Also, set compression=lz4. As noted above under LZ4 compression, larger record sizes will increase compression ratios on compressible data by allowing the compression algorithms to process more data at a time. Furthermore, throughput is generally increased by the use of LZ4 compression. In-compressible data will be stored without compression and writes are so fast that in-compressible data is unlikely to see a performance penalty from the use of LZ4 compression.
Also, set compression=lz4. As noted above under LZ4 compression, larger record sizes will increase compression ratios on compressible data by allowing the compression algorithms to process more data at a time. Furthermore, throughput is generally increased by the use of LZ4 compression. In-compressible data will be stored without compression and writes are so fast that in-compressible data is unlikely to see a performance penalty from the use of LZ4 compression.
Editor
348

edits

Navigation menu