Fun MS bug

dave330i

Ran into this bundle of joy recently:

https://support.microsoft.com/en-us/kb/2769701

Honestly, unless you're trying to squeeze last bit of performance, you probably won't notice the difference.

Deathmage

hrm, I'd be interested to see if this has a positive effect on our database servers running 2008 R2 with a back-end SAN. - I'll follow this, would be interesting to see if this helps.

jibbajabba

Deathmage wrote: »

hrm, I'd be interested to see if this has a positive effect on our database servers running 2008 R2 with a back-end SAN. - I'll follow this, would be interesting to see if this helps.

Seems to affect writes only though ... so might really make .. well nuff all difference for most

dave330i

jibbajabba wrote: »

Seems to affect writes only though ... so might really make .. well nuff all difference for most

When we tested, it also made some difference in read.

Deathmage

dave330i wrote: »

When we tested, it also made some difference in read.

Interesting, what server did you test it on 1st? - was it a noticable difference? - since it's a hotfix I presume its only applied manually and not part of a different update?

Deathmage

@dave330i ...thanks for hotfix, I've noticed the past few days, the server(s) the hotfix was applied too, thier IOPS demands on the Equalogic have reduced by a substaintial degree. Normally they hovered around 150 to 400 iops with our database server peaking to around 550 ios during peak usuage, obviously I was leary of doing any patch to a SQL server so it wasn't applied but our application server I did take a chance cause I had a Veeam Backup from the night before, that server normally hovers around 320 iops steadily and it's now not even pushing 25 IOPS all day Friday!!!!!!

End Users asked me what I did because the responce time on files is blazing fast now...

I have to thank for you this addtion Dave, it's a nice idea for the tool box now.

Here is a picture of a server that normally used to hover around 250 IOPs this time of the night, since it does a planning script to sync databases, if you look at the graph it hasn't even batted an eye....

I think it's performing too good now



Also notice that Veeam backups happen quicker now too by about 15 minutes faster each night on the server(s) affected by the hotfix.

dave330i

Glad to help, but are you sure all of it is from the Hotfix? Our results weren't that dramatic.

Deathmage

dave330i wrote: »

Glad to help, but are you sure all of it is from the Hotfix? Our results weren't that dramatic.

Possibly, I did add two more connections from the iSCSI fabric into the hosts and enabled MPIO across all four links so that could have had a factor too...

Mike7

Reminds me of the robocopy slow performance on 2008 R2.
Big difference after applying hotfix and new switches.

ninjaturtle

Deathmage wrote: »

Possibly, I did add two more connections from the iSCSI fabric into the hosts and enabled MPIO across all four links so that could have had a factor too...

Yea, I'm pretty sure your increase in performance came from your new MPIO configuration. Which Equallogic array are you guys using? We just picked up one a couple months ago. We got a 6510e w/ 48 15K 600GB SAS drives. I've got the 10gig interfaces connected back to a Nexus 5548.