> > Hi, > > I understand that the size of the physical RAM has to be > proportional to the > total harddisk cache size. Supposing I have > > unlimited physical RAM, > > > 1) What is the recommended size of 1 physical harddisk for > each server (each > server can have sda, sdb etc...). The reason I > > am asking this is that I am concern that the bigger a disk > is, the longer > squid needs to get an object out of it. > > I am currently using a U320 SCSI disk of 10KRPM, size 73GB. > It is being > partitioned into 4 smaller partition of 17GB each. > > Performace is within expectation, but I am wondering if > reparttioning it > into smaller sizes would give better yield. On the > > other hand, the disk has a fixed number of head, so would it > even help? > > 2) In term of performance only, is a 100GB harddisk better > (partitioned into > 5 20GB partitioned) or 5 20GB harddisks better. > The multiple spindles solution would be better, as SQUID's performance highly depends on (reduced) seek time(s).
M.