Change the compression algorithm of the raw_extrinsic_metadata table in cassandra
This table is quite big but already pretty well compressed by the lz4 algorithm.
cassandra01 ~ % /opt/cassandra/bin/nodetool -u cassandra -pwf ~/.cassandra/nodetool.passwd tablestats -H swh.raw_extrinsic_metadata
Total number of tables: 1
----------------
Keyspace: swh
Read Count: 1942632880
Read Latency: 0.700641182116201 ms
Write Count: 2425436468
Write Latency: 0.05391835442584761 ms
Pending Flushes: 0
Table: raw_extrinsic_metadata
SSTable count: 21
Old SSTable count: 0
Max SSTable size: 143.435GiB
Space used (live): 342.78 GiB
Space used (total): 342.78 GiB
Space used by snapshots (total): 0 bytes
Off heap memory used (total): 927.76 MiB
SSTable Compression Ratio: 0.23764
Number of partitions (estimate): 76360222
Memtable cell count: 34353
Memtable data size: 151.74 MiB
Memtable off heap memory used: 0 bytes
Memtable switch count: 202
Speculative retries: 9161
Local read count: 554143
Local read latency: 0.420 ms
Local write count: 1368220
Local write latency: 0.047 ms
Local read/write ratio: 0.40501
Pending flushes: 0
Percent repaired: 99.97
Bytes repaired: 1.371TiB
Bytes unrepaired: 437.185MiB
Bytes pending repair: 0B
Bloom filter false positives: 32957
Bloom filter false ratio: 0.00000
Bloom filter space used: 221.4 MiB
Bloom filter off heap memory used: 221.4 MiB
Index summary off heap memory used: 4.27 MiB
Compression metadata off heap memory used: 702.09 MiB
Compacted partition minimum bytes: 125
Compacted partition maximum bytes: 386857368
Compacted partition mean bytes: 11204
Average live cells per slice (last five minutes): 1.1044776119402986
Maximum live cells per slice (last five minutes): 2
Average tombstones per slice (last five minutes): 1.0
Maximum tombstones per slice (last five minutes): 1
Droppable tombstone ratio: 0.03698
Top partitions by size (last update: 2025-02-05T21:09:13Z):
Key Size
swh:1:dir:7d7acd7ac3143e9d8e2cf3083947ec0263439ac3 359.3 MiB
swh:1:dir:d7ed6796193a1b9612656e636c9f8453a3db8a6c 348.57 MiB
swh:1:dir:8fb86f0700d1a5b804323c7ad4314b43d0a5d389 305.36 MiB
swh:1:dir:12ae6e71b9c67abd8a13f1981758d04c69d15397 247.77 MiB
swh:1:dir:3d0aa14c866f32eaae1ee5035a7816e68f159833 211.89 MiB
swh:1:snp:f1935cd0a10d1f804de6ff41ba803e083c3f234e 200.16 MiB
swh:1:dir:c130023e984dd8fb798b35716f05833400875f33 186.54 MiB
swh:1:dir:dc93a0f24881173c22feb7948d4915130f93b969 160.3 MiB
swh:1:snp:66395fba63fba4cf872c04595a6e35fa2f572a0f 143.16 MiB
swh:1:dir:2d36f11f3d5c9e156faf7f3f28ebfa292157be17 142.25 MiB
Perhaps we could have a even better compression ratio with zstd.