On Sat, Nov 01, 2003 at 08:37:42AM +1100, Russell Coker wrote: > On Sat, 1 Nov 2003 07:38, Daniel Erat wrote: > > The server has a SuperMicro Super P4DP6 motherboard with dual Xeon > > 2.4Ghz processors and 4 GB of RAM. One of the two onboard Adaptec > > 7899P SCSI chipsets is being used to control a disk that has the OS, > > and a > > My experience is that the kernel.org kernel performs very badly under > heavy disk IO when you have 4G of RAM. There are patches to fix this, > I can't remember which ones they are (and I couldn't find the patches > on their own when I looked). > > I solved this before by using a SUSE kernel source tree, but a Red Hat > kernel source tree has the same patches.
Thanks! I'm hesistant to switch to the SUSE or Red Hat source trees because of the large number of changes there. If I just want to work around the problem so I can finally get some sleep, do you know if I can pull a gig of memory out of the machine, or will I need to instead or additionally recompile the kernel without 4 GB High Memory Support? In other words, is this some sort of boundary condition that's triggered when the kernel is working near the 32-bit 4 GB limit, or is it a more general problem in the 4 GB High Memory Support option that will crop up regardless of how much physical memory is actually in the server? Dan
