Slow Disk I/O
in [Kernel]
|
Prev: [tip:perf/urgent] tracing, powerpc: Fix for tracepoint API change
Next: sched: add wait_for_completion_killable_timeout
From: Brian D. McGrew on 28 May 2010 14:40 Good morning All! I�m running CentOS 5.4 x86_64 on a Dell T5500 with 3GB || 12GB or RAM. I�m running a 2.6.18-128.el5 kernel. There is a 500GB SATA drive connected to the onboard SATA controller. I can reproduce this problem with CentOS 5.4 i386 on the Dell T5500, T5400 and Optiplex 740�s. So now, on to the problem... In our software, we�re reading in 5MP image files in the neighborhood of 500 to 1500 files at a time. Just a simple for() loop, open, read, close... Nothing fancy... When I have 3GB of RAM in the system, life is good.. Each read takes 8 to 10ms... This is a good thing! If I bump the memory up to 12GB, all of the reads are now taking 150 to 200ms. My default, va.swappines to set to 60... If I decrease this number, the problem gets much worse. If I up va.swappines to 100, the problem gets a little better, but not a whole lot. What should I be looking at??? -b -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo(a)vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
From: Robert Hancock on 28 May 2010 20:10
On 05/28/2010 12:34 PM, Brian D. McGrew wrote: > Good morning All! > > I�m running CentOS 5.4 x86_64 on a Dell T5500 with 3GB || 12GB or RAM. I�m > running a 2.6.18-128.el5 kernel. There is a 500GB SATA drive connected to > the onboard SATA controller. I can reproduce this problem with CentOS 5.4 > i386 on the Dell T5500, T5400 and Optiplex 740�s. > > So now, on to the problem... In our software, we�re reading in 5MP image > files in the neighborhood of 500 to 1500 files at a time. Just a simple > for() loop, open, read, close... Nothing fancy... > > When I have 3GB of RAM in the system, life is good.. Each read takes 8 to > 10ms... This is a good thing! > > If I bump the memory up to 12GB, all of the reads are now taking 150 to > 200ms. > > My default, va.swappines to set to 60... If I decrease this number, the > problem gets much worse. If I up va.swappines to 100, the problem gets a > little better, but not a whole lot. > > What should I be looking at??? Can you post the dmesg output and the contents of /proc/mtrr? -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo(a)vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/ |