[Lessfs-users] Very slow performance lessfs 1.5.12 with hamsterdb 2.0.4.
Brought to you by:
mruijter
From: Dominic R. <do...@ti...> - 2012-07-26 10:19:31
|
I am experimenting with lessfs on Ubuntu 12.04 using the configuration suggested by Mark at http://www.lessfs.com/wordpress/?p=707, but with lessfs 1.5.12 and hamsterdb 2.0.4 (and snappy 1.0.5). I have used Mark's lessfs.cfg file (http://www.lessfs.com/wordpress/wp-content/lessfs.cfg) but as I am testing with a single-processor VM I set MAX_THREADS=1. The machine has 1GB RAM. I set about copying (rsync) some data (a lot of files, mostly quite small Word or pdf files) into the lessfs mount. Things go quickly for a while (and there are pauses every so often which I guess are when lessfs flushes data to disk) but then become very slow indeed (say 1 x 30K file per second), and never speed up again. This seems similar to the problems in a previous posting (https://groups.google.com/forum/?fromgroups#!topic/lessfs/cA2M6Czc3Us) but that was where tokyocabinet was being used, and the suggestions relate to _BS configuration options which don't seem apply when using hamsterdb: man lessfs states: 'The hamsterdb database does not require the _BS lines since hamsterdb does not use hash databases'. Does this apply to all *_BS configuration options, I presume so because none are set in Mark's lessfs.cfg? Size of underlying datastore (sudo du -h --summarize /data): 589MB Size of stored files (sudo du -h --summarize /mnt/less): 869M (62 min to calculate this!!!) Number of stored files (sudo find /mnt/less -type f | wc -l): 22675 Can anyone suggest why lessfs is so very slow on my system and how I can improve it? I have tried starting lessfs with '-o big_writes,max_read=131072,max_write=131072', or not: doesn't seem to make any difference to performance, maybe the setting BLKSIZE=131072 in lessfs.cfg means these settings are used anyway? Thanks for any help Dominic -- *TimeDicer* <http://www.timedicer.co.uk>: Free File Recovery from Whenever |