I prefer well-designed caching policies ( refresh_pattern ) to complicated caching hierarchy. Simply loadbalancer - squid caching servers (with no sibling relationship) - original servers. Possibilities are that when you get one cache miss on one squid server you get the same cache miss on the other caching servers. > -----Original Message----- > From: Paras Fadte [mailto:plfgoa@xxxxxxxxx] > Sent: 2008年8月12日 0:57 > To: squid-users@xxxxxxxxxxxxxxx > Subject: Squid Caching Hierarchy > > Hi, > > In the following S1,S2,S3,S4,S5 represent squid caching servers. A > request coming through Load balancer will go to S1,S2 or S3 and if > they don't have the requested web object it will contact S4/S5 which > In turn would contact O1 or O2 which are origin servers. Is this setup > plausible enough to be efficient or is there some other hierarchy > which could be more effective ? The aim is to keep load on origin > servers to the minimum and also the latency to minimum. > > Load Balancer > > S1 S1 S3 S4 S5 > > O1 O2 > > Thank you. > > > -Paras
