Chris Robertson wrote:
Everything I've read says that you should not use any RAID for your cache directories. Make a bunch of cache dirs (each on its own disk), and let Squid sort it out.
Chris
I just want to add that 250GB of cache needs a lot of memory for the index. The figure I know is 10M of RAM for 1GB of cache.
It is a good thing to use multiple separate drive to spread the i/o but I would carefully increase the cache size while monitoring the memory usage.
I have recently switched vom striping to separate drives. The performance is similar, but if one drive fails only one cache_dir is lost. With a stripe you will lose everything. (Thanks to Adam Aube for pointing this out for me).
Regards, Hendrik Voigtl�nder
-----Original Message----- From: J Thomas Hancock [mailto:[EMAIL PROTECTED] Sent: Monday, November 29, 2004 10:50 AM To: [EMAIL PROTECTED] Subject: [squid-users] Disk Configuration
I am currently configuring a small army of squid servers to act as transparent caches. They are running Fedora Core 3 and Squid 2.5. We are using disk striping to have a fairly large and fast cache directory.
I have heard in older versions of Squid, that there were performance issues if the cache directory got to large. We currently have approximately 250GB of storage for our cache directory. Would you recommend having one huge 250GB cache directory or would you recommend having several smaller cache directories all on the same disk?
Thank you, Tom
