On Fri, Aug 17, 2007 at 11:07:14AM -0400, Andy Greenwood wrote: > My understanding was that you should estimate swap size based on the > sizes of the programs which might be paged out. However, when I first > set up my system, I didn't know this and created 1G swap slices (one on > each disk) but I am not convinced that this was the best thing to do, > since my system almost never uses a noticible percentage of the swap > space. right now, I've got > > [EMAIL PROTECTED] fusefs-sshfs]$ swapinfo > Device 1K-blocks Used Avail Capacity > /dev/ad0s1b.eli 1048576 1148 1047428 0% > /dev/ad1s1b.eli 1048576 1096 1047480 0% > Total 2097152 2244 2094908 0% > > And the system is under normal load. This system has 1G of RAM. Is there > any sense in having this much swap space when it's not being used?
swap is there to guard against overload conditions, not for normal load. If you are paging during normal operations your system performance will be terrible, so you want to make sure you have enough RAM that this does not happen. However, when a transient load spike comes in, would you prefer your system to slow down but keep working, or to kill off all your processes? Think of it as memory space insurance. Kris
pgpvcvTfETSpH.pgp
Description: PGP signature
