Subject: Re: ccd performance
To: None <current-users@NetBSD.ORG>
From: Dave Huang <>
List: current-users
Date: 06/13/1997 17:35:24
I recently made my /usr a ccd (just for fun and 'cuz I think disk striping
is nifty :) and was wondering how to determine the best interleave
setting. Is the size of a track still a good guideline to use? My disks
aren't identical--they're not even the same speed, which is probably
especially evil.

One's a 5400rpm Quantum Capella and the other's a 7200rp Quantum Atlas.
Both are use ZBR (or whatever that thing's called where outer tracks have
more sectors than the inner ones), but according to the messages when the
drives are probed, one's got 126 sectors/track and the other's got 115. 

I tried a ccd with an interleave of 127, and untarred the 600 megs of
stuff that used to be on the drives before I made a ccd, and often there
were rather lengthy periods (5s or more) where only one drive would be
active. So, I reduced the interleave to 32 and did the untar again, and
both drives seemed to run simultaneously more, although there were still
some short periods where only one drive was active. I didn't actually try
any benchmarks to see which way was better, but I figured that running
both drives simultaneously was probably better :)

On Fri, 13 Jun 1997, Jason Thorpe wrote:
> If you find your freelist misses are high, you can set the kernel config
> option "CCDNBUF" to some value.  That value indicates the number of
> ccdbuf's that will be allocated per component.  The default is 8 (which,
> in your case, means you have 2 components, since your hiwater is 16).

Is "high" an absolute number (e.g. more than x is high), or relative
compared to the number of ccdbuf allocations (e.g. over y% is high)?

Name: Dave Huang     |   Mammal, mammal / their names are called /
INet:   |   they raise a paw / the bat, the cat /
FurryMUCK: Dahan     |   dolphin and dog / koala bear and hog -- TMBG
Dahan: Hani G Y+C 21 Y++ L+++ W- C++ T++ A+ E+ S++ V++ F- Q+++ P+ B+ PA+ PL++