fre 2010-05-21 klockan 06:38 +0000 skrev GIGO .: > can it be said as a generalization that one can allocate/fix 1/4 of > physical ram for cache mem objects. Will it holds true even when you > are running multiple instances??? I would not generalize a rule like that. It is a reasonable recommendation when sizing the system, but also depends on how your Squid is being used. A reverse proxy benefits much more from cache_mem than a normal forward proxy, and in a forward proxy you may want to give priority to on-disk cache instead. memory usage per Squid = cache size (in GB) * 10 MB + cache_mem + 10MB. memory usage by OS: Leave at least 25%. In smaller configurations up to 50%. system memory requirement = sum(squid instances) + system memory = sum(squid instances) / 0.75. If you inverse the above calculation then you'll notice that cache size is a function of cache_mem. If one is increased then the other need to be decreased. Note: if you also log in on the sever using graphical desktop (not recommended) then reserve about 1GB for that. > please guide that how memory handling will be occuring in multiple > instances setup???cache_mem will influencing per instance and not the > program as whole. right? Right. Regards Henrik
