HPE Storage Users Group
https://www.3parug.com/

VLUNs Performance Statistics
https://www.3parug.com/viewtopic.php?f=18&t=2811
Page 1 of 1

Author:  nishanth [ Fri Mar 02, 2018 12:02 am ]
Post subject:  VLUNs Performance Statistics

Hi Team,

Currently we are getting lot of concerns related to performance issues regarding applications running on VM's and I can see those LUN's are made up of Balanced AO, where all three disks are part of it. SSD, FC10K and NL. By default data goes to FC10K and majority of the data also is sitting on FC10K too.

When I ran report of VLUNs - Performance Statistics by selecting IOPs, Bandwidth, Service Time, I/O Size, Queue Length and Average Busy, sometimes the service time is above 30ms and Queue length is above 25 also. So would like to know what is the ideal values for this to be with in the threshold for best performance as these LUNs are residing in multiple type of disks. Can someone help me in understanding whether any issue from Storage side which is causing the sudden spike in these values. Also I noticed that though all these values being high, I can still see that the average busy chart is 0 only. I searched online a lot regarding the ideal values for these parameters but couldn't get a satisfying reply from anywhere. some where i saw that queue length more than 2 is something to be worry about and sometimes i can see in my environment where queue length is more than 25. Also somewhere it is mentioned service time above 20 ms is to be worried about and I can see it is more than 30 ms for some of my VLUN's. Any help on this will be much appreciated.

Author:  Patrick [ Fri Mar 02, 2018 11:58 am ]
Post subject:  Re: VLUNs Performance Statistics

I'd take a look at which VM's are generating the most I/O then move them over to a separate SSD/10K CPG. If you have any workloads with a heavy write pattern you could also try separate CPG for that with Raid1 on the 7k tier.

Also make sure your block sizes within the guest os are correct for the workload, i've seen a couple places forget to use 64k blocks for SQL and video storage causing orders of magnitude more unnecessary IO.

Author:  nishanth [ Sun Mar 04, 2018 11:01 pm ]
Post subject:  Re: VLUNs Performance Statistics

Patrick wrote:
I'd take a look at which VM's are generating the most I/O then move them over to a separate SSD/10K CPG. If you have any workloads with a heavy write pattern you could also try separate CPG for that with Raid1 on the 7k tier.

Also make sure your block sizes within the guest os are correct for the workload, i've seen a couple places forget to use 64k blocks for SQL and video storage causing orders of magnitude more unnecessary IO.



Thank you for the input Patrick, for one application related LUN I can see write request is more as you mentioned.

Page 1 of 1 All times are UTC - 5 hours
Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group
http://www.phpbb.com/