I/O on SSD hosting a DB periodically hangs with queue length peaks of 800+

performancessd

i have a problem which i cannot identify correctly. I have a process (on WinSrv 2008 R2) which imports data into Mongo DB and periodically whole system freezes – once per 5-10-15 minutes for about 20-30 seconds. During freezes HDD led on the computer box is constantly 'on' and all processes that somehow access SSD are not responsive. After some time (15-30 seconds) system unfreezes and perfmon shows disk queue length of 800+ (whereas normally it is not more than 5-10).

Screnshot from perfmon is here http://clip2net.com/s/5wVBPP

I suspect there is some hardware problem with SSD, which is Intel 520 120GB (was tried with similar or worse result /system hangs completely/ – in different boxes with latest SSD/BIOS firmwares and drivers). The question is – are there are some other reasons possible?. Is it normal for Intel 520 /my company installs only Intel hardware, so there seems to be no other hardware choice/?

Best Answer

Hardware or Software issue, with Software being the firmware on the drive.

The SSD likely decides it needs to do some cleanup / garbage collection RIGHT NOW and stops any answering until that is done. CHeck firmware versions, apply upgrade as feasible, otherwise RMA drive for replacement.

Not sure about Intel - they have a REALLY good reputation - but OCZ did that in older versions at times, though they got that under control finally ;)

Related Topic