tis 2010-03-30 klockan 22:48 -0300 skrev Alex Montoanelli: > My question is, Can squid slow down, when used with large cache_dir, > and they are close to 80% of use? I/O load on the cache dirs gets much more intensive when Squid starts to recycle space. This is due to most filesystems not being very good at removing data and performs about as much work when removing data as when storing the data in the first place. This is the most common cause to slowdown when the cache has been filled. You can use system tools such as sar, iostat etc to monitor your system components to determine if this is a likely cause, seen by large amounts of I/O wait cpu state & I/O transactions to the harddrive. The second cause on the common list is shortage of memory. The larger your cache is the more memory Squid uses, and when Squid approaches the maximum memory usage your system can support it then starts to slow down rapidly. Obvious signs of this is swap space usage starting to increase. Third it's also known that there is some internal issues in Squid, especially seen in very large memory configurations, where CPU usage from time to time spikes and performance drops. A workaround which seems to help is setting "memory_pools off" in squid.conf. Regards Henrik
