On Tue, 9 Mar 2004 [EMAIL PROTECTED] wrote: > I'm looking at moving our Squid servers, where cache is currently > configured as 4 x 9G on 4 separate disks, to new hardware with 2 x 18G > disks set aside for cache. > > Is it preferable, from a performance perspective, to configure cache on the > new platform as 2 x 18G areas or 4 x 9G? Or no difference?
More spindles is almost always preferrable, but it also depends on the quality of the spindles.. Squid eats seeek time. For this reason you want the sum(drive_seek_time / number_of_drives) / number_of_drives to be as small. Lets assume one configuration has 4 drives with 5 ms seek time, on the other 2 drives with 4 ms seek time. Their relative "slowness" (smaller is better) is then 4 x 5ms: ((5 / 4) * 4) / 4 = 5 / 4 = 1.2 2 x 4ms: ((4 / 2) * 2) / 2 = 4 / 2 = 2 Regards Henrik