10GB Performance Disaster

[1parkplace]

interesting... Those are the settings I use and all works for me. not sure why they wouldnt work for you. I was originally given those settings by an Open-E tech.

I switched to Block I/O recently, fixed a lot of problems for me. This was from advice from Open-E, they said they certified DSS with VMWare with Block I/O, not File I/O.

I deleted all my volumes and recreated as Block I/O and then VMWare saw all my LUNs and the performance was still on par.

[Robotbeat]

interesting... Those are the settings I use and all works for me. not sure why they wouldnt work for you. I was originally given those settings by an Open-E tech.

I switched to Block I/O recently, fixed a lot of problems for me. This was from advice from Open-E, they said they certified DSS with VMWare with Block I/O, not File I/O.

I deleted all my volumes and recreated as Block I/O and then VMWare saw all my LUNs and the performance was still on par.

Block I/O is better. Fewer caching layers (I think), lower latency. In some circumstances, that means cached I/O can be a little slower, but if you have a fast back-end (a decent RAID card), it works fine. In fact, in some cases, File I/O just gets in the way.

[1parkplace]

Block I/O is better. Fewer caching layers (I think), lower latency. In some circumstances, that means cached I/O can be a little slower, but if you have a fast back-end (a decent RAID card), it works fine. In fact, in some cases, File I/O just gets in the way.

I agree. Believe it or not... my machines have a great hardware backend and 32GB of RAM each, but File I/O just couldn't cut it.

I have a simple setup with 2 ESX servers and 2 DSS servers, each ESX targets both DSS and has an even split of VMFS volumes on each DSS.

After switching to Block I/O... Storage vMotion, vMotion, etc all sped up greatly(10x faster in some cases), barely increased "High Disk Latency" counter on the ESX servers during storage moves.

Overall I am much happier using Block I/O.

[symm]

Hey 1parkplace

before you changed to block I/O did you get and cmd_abort (1143) errors?
did going to Block I/O fix this?

[1parkplace]

Hey 1parkplace

before you changed to block I/O did you get and cmd_abort (1143) errors?
did going to Block I/O fix this?

Not 100% sure, I know I used to get those before DSS v6 and also with the NetXen 10GB cards.

The upgrade to v6 and the intel 10gb happened at the same time, so not sure which fixed my problems.

I would assume it would probably remedy that situation though, since it takes caching in RAM out of the equation. As long as you have the spindles to keep up with the data load and your RAID controller is solid, you shouldn't have those problems.

[Alex1411]

Not 100% sure, I know I used to get those before DSS v6 and also with the NetXen 10GB cards.

The upgrade to v6 and the intel 10gb happened at the same time, so not sure which fixed my problems.

I would assume it would probably remedy that situation though, since it takes caching in RAM out of the equation. As long as you have the spindles to keep up with the data load and your RAID controller is solid, you shouldn't have those problems.

Thanks for answer!