0
My C:\pagefile.sys file is currently sitting at 107,827 ms response time. Resource Monitor reports that 74% of memory is in use. If I leave the computer with minimal interaction for a while (such as while typing this question) the response time starts to lower (I assume it is a trailing average?). When I launch a new application its process goes through a few (dozen) seconds of hard faulting, and the pagefile.sys response time jumps back up.
Additionally, sometimes C:\$Mft (NTFS Master File Table) and C:\$LogFile (NTFS Volume Log) have very high response times too, e.g. $Mft is currently being reported at 1,333,520 ms and 1,000,140 ms (it's alternating between the two).
Performance Monitor reports a current disk queue length of 0.00 last, 0.020 avg and 1.000 max and an average disk queue length of 1.627 last, 2.190 avg and 26.505 max.
What might be causing this and how should I approach troubleshooting it?
I've read related questions such as Windows 8.1: pagefile.sys disk response time > 1S but haven't found an answer.
2a) I"ve seen defective drives to this. You're looking at retries. b) Having a long request queue to the drive can do this. Is it on a drive by itself? Can you try putting it on a drive by itself? c) On some recent motherboards, the Intel SATA controller and its driver have been blamed - fixed by updating the RST driver (even if you're not using your drives in RST mode). – Jamie Hanrahan – 2017-04-18T00:23:40.503
Also, please clarify - do you mean about 108 milliseconds, or about 108 seconds? (the comma is ambiguous) And: Have you checked the System error log for disk errors? – Jamie Hanrahan – 2017-04-18T00:46:52.383
I mean about 108 seconds - the value is as shown in Resource Monitor, which in my locale uses a comma as a thousands separator. – Kendall Lister – 2017-04-18T00:57:48.887
I don't have any other drives in this machine, so I can't move the pagefile.sys to another drive, but I should probably have mentioned that this machine does have two drives in RAID with Intel Rapid Storage Technology. It is an old motherboard though (maybe 6 years) and I have been keeping the drivers up to date. I'll check the queue length, and the system error log - thanks! – Kendall Lister – 2017-04-18T01:01:11.050
There are no obvious disk errors in the System log. I've added queue length details to the question. How can I inspect the retries? – Kendall Lister – 2017-04-18T02:35:05.397
asaik retries are transparent to a modern OS. If I were you I would break your mirror set and try running with just one drive. – Jamie Hanrahan – 2017-04-18T03:25:16.477
Can you point me to a good guide for doing that safely? Is the answer to this question a good method: https://superuser.com/questions/137308/removing-raid-1-mirroring-and-leaving-data-on-both-drives ?
– Kendall Lister – 2017-04-18T03:31:20.220Here's a thread at the intel support forum, linking to a procedure, with an "it worked!" followup. https://communities.intel.com/thread/48706
– Jamie Hanrahan – 2017-04-18T07:44:30.180Yes, I saw that page - you don't have any first-hand experience? I'll give those instructions a go when I have time. – Kendall Lister – 2017-04-18T08:23:12.390
exp with lots of RAID stuff but not RST. – Jamie Hanrahan – 2017-04-18T13:24:28.777