Hi Guys, hope everything is alright, I'm writing here because I have observed a curious behaviour in our VNX system. We have a VNX 5700 with 3 Data Movers, 2 active, 1 stand by. And basically what I'm seeing is one of the DMs is hitting 1,800 IOPS top while the other one is hitting between 2,000 and 12,000 IOPS. This behaviour is given due to the type of shares we host in each of the, the one with fewest IOPs hosts mostly user shares, while the other one hosts application shares. The question is mainly because I haven't been able to find any benchmark doc or any document describing the limitations.
I have checked all the other VNX systems we have in the company and all of them are around 1,800 CIFS IOPS.
Couple of weeks ago we migrated an application which has intense IO to the DM hitting 12,000 IOPS, we had to roll back since the while the application was writting to the files, it dropped the connections. We are still not quite sure the problem is related to VNX system. But wanted to confirm with the community if there is a documented limit of IOPS in what regards to CFIS.
Even we are seeing lots of IOPS we are not seeing an increase in the netwrok utilization, so probably we could fix that by enabling jumbo frames. But anyway IOPS is still a concern
Thnk you guys in advance for your answers!!
not all IOPS are equal
some are small and light metadata read operations that can be done from the cache – others are large read and write that hit the disks
I don’t think we drop connections – unless they are idle or there are network problems
Thanks for your reply! Is there any benchmark document available for the VNX5700 Unified or File? I'd like to see if we are actually stressing the box, because so far I can tell one of the DMs has more IOPS than the other, but can't tell if it actually is a problem.