Comments on: Lessfs-1.1.9.1 has been released.

By: Mike

Mike — Mon, 04 Oct 2010 21:06:21 +0000

Thanks for your quick reply. As for the IOPS, that makes me feel better about my grasp of the subject. As for the configuration settings, I am refering to the tuning required by Tokyo Cabinet. I have a working config and now I just want to be sure that I am running as fast as my system can.

By: maru

maru — Mon, 04 Oct 2010 17:38:58 +0000

Hi Mike,

My current laptop gives me these numbers:
root@saturn:/usr/src/writetest# ./writetest
running: 1000 ops
Elapsed time: 30.994890 seconds
1000 ops: 32.26 ops per second

The 6400 IOPS come from a HP DL360 with a CCISS controller that comes with a reasonable large write cache.

The current LFS 1.1 code relies less on IOPS then the LFS1.0 series did.
My laptop can sustain approx 40~60MB/Sec with LFS. With a fast processor and a slow disk you can try a high compression algorithm like bzip. This will reduce IO at the cost of processor load.

If you want to create a properly configured lessfs.cfg file without going through the manual try:
http://stashbox.org/900747/lessfsconfig-0.0.2.tgz

I hope this helps,

Mark

By: Mike

Mike — Mon, 04 Oct 2010 16:28:15 +0000

I tried lessfs again and have a few questions about performance tuning. Most of the Tokyo Cabinet stuff is above my head and I haven’t finished wandering through the help file for it. I read a post about another user having issues with copying files into his lessfs directory and the steps he took to troubleshoot. One thing he tried was running a compiled c file called writetest.c to determine his ops. I noticed that he got about 6400 ops. I have compiled and ran it several times but my ops are between 30-50 with is quite low. Does ops refer to IOPS? If so, then I have read that SATA disks only can handle around 50-75 IOPS in a single disk setup and you have to use raided SSD before you hit anything like 6400. Sorry if I’m barking up the wrong tree, but I would really like to try lessfs out!