PowerPath VE Versus Round Robin on VMAX – Round 2

October 11th, 2011 |

As promised in the first post, here is round 2 of my testing with PowerPath VE and vSphere 5 NMP Round Robin on VMAX. For this round of testing, I changed the Round Robin iooperationslimit to 1, from the default of 1000.

I understand that this is not recommended, and I also understand that further testing is needed with multiple hosts, multiple VM’s and multiple LUN’s. As soon as I get the time, I’ll do that and report back.

For the background, and methodology, click the link above to read the first post. For now, I’ll skip right to the scorecard.

As we can see here, setting Round Robin IOPS to 1 definitely evens the score with PowerPath. I expected to see more CPU activity than PP, but that wasn’t the case. I also expect to see more overhead on the array once I add more hosts, VM’s and LUN’s to the mix. It might be a few weeks before I can pull that off.

Thanks for reading, and commenting. Round 3 to come.

FredK

Excellent post !
Impressive the difference between the RR config. from 1000 to 1 io operations limit…

So to conclude… powerpath is not really worth the price, isn’t it ?

FredK
- http://twitter.com/BrandonJRiley B. Riley
  
  Fred:
  
  I can’t say that yet. For a single host running one VM from one LUN, it’s not worth the price.
  
  My next round of tests will be multiple hosts, VM’s and LUN’s. I will report back when those tests are complete. It’ll take me a few weeks.
  
  Stay tuned!
  - http://blog.cowger.us Matt Cowger
    
    Exactly – under a 0 contention, perfectly symmetric workload like what is being done here, PP/VE and NMP/RR will behave nearly identically, as we can see.
    
    Once you get a workload that begins to push the array or FAs to their limits, or with mixed workloads (mixed IO sizes and IO profiles), you’ll see PP/VE pull away again.
http://twitter.com/BrandonJRiley B. Riley

Fred:

I can’t say that yet. For a single host running one VM from one LUN, it’s not worth the price.

My next round of tests will be multiple hosts, VM’s and LUN’s. I will report back when those tests are complete. It’ll take me a few weeks.

Stay tuned!
Daniel

EMC engineer here, so I am an interested party. I suspect that PowerPath would regain some of its lead over RR w/ IO/path set to 1 with a more mixed workload. That is, along with your 4k and 8k random operations, throw in a backup workload (really big sequential reads and writes), or maybe something that looks more like DSS (big random reads). I’ve seen similar gains to your Round 1 post at customers w/ real world workloads, even when setting that RR value very low in NMP.

This is largely because w/ mixed workloads from lots of hosts, you can encounter not-insignificant serialization delay (ie, lots of small IOs from some OLTP database backing up behind a few large IOs from other processes), as well as some interesting adverse effects from piling up different layers of queues (in the host OS, in the HBA, in the SAN, on the FA processors). PowerPath’s control scheme does a very good job of mitigating these issues.
- http://twitter.com/BrandonJRiley B. Riley
  
  Daniel:
  
  Do you have IOmeter templates, or settings for those types of tests? I’m setting up a few hosts right now to do a better comparison.
  
  Thanks for the feedback.
Pingback: Nothing found for Index Php 2011 11 09 Powerpath-ve-versus-round-robin-on-vmax--round-3-tko

PowerPath VE Versus Round Robin on VMAX – Round 2

Related Posts: