> Can you give me, please, some hints on fine-tuning XFS for better
> performance for web caches?
> A typical Squid cache looks like this:
>
> - HDD size: 10...30 GB
> - average file size: 10...15 kB
> - 16 primary directories
> - each primary dir contains 256 secondary directories
> - each secondary dir can contain up to 256 actual files
> - the actual disk usage when the cache is "full" is ~80% (can be tuned
> from squid.conf)
>
> >From the FS point of view, the access to the files is quite random. The
> cache fills up first to the maximum size, and then the oldest files are
> deleted and replaced with newer files as the clients request them.
>
> At this point, the only option i use is "noatime".
> Do you have any suggestions to improve the speed?
Using -o logbufs=8 may help.
There is a perl script which is only in the cvs tree at
cmd/xfsmisc/xfs_stats.pl
If you run this script then it will produce a couple of pages of statistics,
there is a section called Log Operations near the end of the first column, if
xs_log_noiclogs is incrementing rapidly then the logbufs option will help.
If you install pcp (and being internal to sgi you can get access to the
graphical interface to this) you can look at these numbers in realtime.
It would be interesting to see the rate at which the Directory Operations
statistics move, these are where lookup, create and remove show up.
In the Tail-Pushing Stats at the top of the second column the xs_sleep_logspace
counter going up says that you are driving the log fairly hard and may benefit
from a bigger log.
I am not sure if you will benefit from using a larger inode size in this
filesystem.
Steve
>
> --
> Florin Andrei
|