Unsolved
This post is more than 5 years old
27 Posts
0
729
What are the best practices for distributing workload on the VMAXe
I can see in WLA that the IOPS are sometimes touching 1900 , we were hit up with a preformance issue a day back.
FAST VP was active moving extents between tiers to balance the load , but still there was the performance issue.
I suspect the workload is not getting balanced.
Can anyone point me to be best practices for distributing the load across the VMAX and how can i point that the performance degradation was primarily due to the high IOPS thrown by the server also how to take care of such spikes ?
Thanks in advance.
Quincy561
1.3K Posts
1
April 9th, 2012 08:00
1,900 isn't a lot of IOPs, unless the IO size is very large, like 1MB, where that would be almost 2GB/sec.
I would start by making sure you are using enough front-end CPUs before you worry about the backend. VMAXe should have the workload spread over the backend by default, but the front-end is up to the customer.
Quincy561
1.3K Posts
0
April 9th, 2012 09:00
2 engines should have a total of 16 available FA CPUs, and it sounds like you are using 4.
I think I would start with SPA or STP data and try and see what components are most utilized, if you think the storage is a bottleneck.
debmdig
27 Posts
0
April 9th, 2012 09:00
The write Kbytes/sec is recorded max at 46000 and the read Kbytes/sec 75000.
The hosts are zones to 4 FAs ( 1f:0, 2F:0, 3F:0, 4F:0 ).
The Storage has 2 engines. This is the max possible under the current config.
Will it be feasible to increase the number of engines by one.
debmdig
27 Posts
0
April 9th, 2012 12:00
In the heat map of the SPA , i was able to see the FA's and the disks to be red.