Experience working with customer escalations suggests that disabling
compression improves performance, and the storage overhead is generally
not a concern for the metadata and omap data we are storing.
Signed-off-by: Sage Weil <sage@redhat.com>
(cherry picked from commit
b878ead071b328e5fe7309a2368383e67679e9f7)
Conflicts:
src/common/options.cc
Config options moved files since jewel.
OPTION(rocksdb_collect_memory_stats, OPT_BOOL, false) //For rocksdb, this behavior will be an overhead of 5%~10%, collected only rocksdb_perf is enabled.
// rocksdb options that will be used for omap(if omap_backend is rocksdb)
-OPTION(filestore_rocksdb_options, OPT_STR, "max_background_compactions=8;compaction_readahead_size=2097152")
+OPTION(filestore_rocksdb_options, OPT_STR, "max_background_compactions=8;compaction_readahead_size=2097152;compression=kNoCompression")
// rocksdb options that will be used in monstore
OPTION(mon_rocksdb_options, OPT_STR, "write_buffer_size=33554432,compression=kNoCompression")