Array Performance and Data Protection
cancel
Showing results for 
Search instead for 
Did you mean: 

Hyper-V 2012R2 Cluster Through SMB 3.0 Shares

 
SOLVED
Go to solution
ndyer39
HPE Blogger

Re: Hyper-V 2012R2 Cluster Through SMB 3.0 Shares

*Edit - i wrote this last night but I can see it didn't apply, sorry*

Hi David,


Thanks for the graphs! As you're probably aware sequential IO on nimble is derived from cpu and nl-sas, rather than SSD.


I can see you're running cs200's. These controllers are great for random IO but get swamped on heavily sequential read/write workloads, which I think is what you're seeing here. This would also be reflected in Infosight in the Performance->Details page under CPU Saturation and Sequential IO Contention. You would see higher performance (lower latency and higher mb/sec throughput) with the larger controllers.

Nick Dyer
Nimble Field CTO & Evangelist

twitter: @nick_dyer_
dejvnull49
Advisor

Re: Hyper-V 2012R2 Cluster Through SMB 3.0 Shares

Hi Nick!

Yeah, that's what we just got, it just took very lookg for the support guys to get to us on that this was the fact.

The MS guys seem very lost on just this and are pushing to replace Nimble with a Storage Space/SOFS solution using a JBOD with SSD and SAS.

The thing is that it is not a heavy sequential workload, we did only the test above and well to put it plainly, the unit really sucks at this type of workload.

This is the workload from a Windows Copy job...

cpu-load-cs240g.png

At the moment I have to look in to getting the controllers swapped and see if the cost will be really bad or not.

It's that or have them used behind a SOFS or similar for other jobs.

/Dave