Please don't post HTML mail to the list - set your mail client to send in plain text.
Really sorry..
>i have go through a lot of document and spent a lot of time browsing to >find the truth about Page faults with physical i/o: but still could not >find the solid answer and solution and all this very confusing.
I assume you're looking in Cache Manager's "General Runtime Information" screen. The "Page faults with physical i/o" statistic is the number of times Squid looked for a requested object in memory, didn't find it, and had to load the object from disk.
Divide the number of page faults by the "Number of HTTP requests received" statistic, then multiply by 100 to get a percentage. Let us know what this percentage is - I honestly couldn't say at what level one should be concerned.
Disk is slower than memory, so if the number is really high then it can impact performance. The solution is to allocate more memory for Squid to hold cache objects in (the cache_mem parameter in squid. conf).
Thanks a lot for useful info. Yes that info. provided are from Cache Manager "General Runtime Information". FYI, the percentage is Page faults with physical i/o / Number of HTTP requests received is 0.29%. The physical memory is 256 DDR, and the swap cache dir is 5000 16 256. I use aufs, heap lfuda replacement policy.
Please Advice..
