Unfortunately this is not something we can do very much about. I run into this problem ALL THE TIME, and we don't have any decent solution. There are 2 issues here: the read cache, and the write cache.
You can "get rid" of the write cache by using the mount option -o sync but that will result in eye-popping performance reductions. I tried that, wasn't worth the performance hit. There's really not much way of "tuning" the Linux buffer cache behavior. That's why the big DB vendors etc. use raw filesystem, to get around the O/S caching mechanisms and gain more control. I've even seen this behavior (overly greedy buffer cache starves applications of RAM) even on AIX. But then I don't know enough about AIX to know if it can be worked around on that platform. Another thing you can try is run with no swap... On Dec 15, 2007 7:12 PM, Kelsey Hartigan Go <[EMAIL PROTECTED]> wrote: > Hi. need help here. > Using RHEL 4 update 5 with 4 GB of memory. (kernel 2.6.xx) > Have an application that is used by maybe 80 to 100 users at a time, but > disk cache doesn't seem to give way to the application and at times, the > system will swap in/out heavily, which blocks the running processes, even > though there's still 2.5 to 3 G in the disk cache. > > tried swappiness = 0 but no dice. > > is it possible that the apps is limiting it self to use a smaller memory? > or is it possible that the disk cache does not shrink because the pages are > always dirty? > > Long term and expensive solution is to add more memory, but before that, any > insights will be helpful. _________________________________________________ Philippine Linux Users' Group (PLUG) Mailing List [email protected] (#PLUG @ irc.free.net.ph) Read the Guidelines: http://linux.org.ph/lists Searchable Archives: http://archives.free.net.ph
