Mixing 3G and 6G SAS drives - Performance Metrics

[sfcredfox]

Community,

I'm interested in doing a performance test against a 3G/6G SAS pool. I'd like some suggestions on the *best* (subjective, I know) or better ways of doing performance comparisons to get more usable results when using a VMware VMFS workload of around 10 typical windows VMs.

I currently have two disk pools (1) 24 disk 3G SAS pool, and (2) 12 disk 6G SAS pool. Each pool has an SSD SLOG and cache drive, and are mirrored pairs.

I am curious is anyone has personal performance testing two pools like that in a way that lends more usable results?

I know I could run a DD on the two pools as they are now (separated), and then run a DD on the pool after combining them, but what would be the best set of test parameters you'd use for VMware VMFS5?

The overall question I'm trying to answer is: can similar performance be achieved by combining the 36 drives into one pool of mirrors, or do they have to be kept separate to achieve reasonable results?

In principal, combining the drives into one pool should just have them all operating at 3G speeds. my hunch is the 6G SAS drives don't get 6G speeds anyway, but I need a better way to quantify that.

Any ideas?

 

[nightshade00013]

The slowest part of the system will be how fast the drive can read from the platter's. The fastest HDD out there can't saturate the SAS 3G connection. All things being equal (rotation speed, capacity, density, cache, etc) the drives will all perform similarly.

 

[Johnnie Black]

You could lose some performance if using a SAS expander connected to disks with different linking speeds, as it will bring down the expander linking speed when accessing them simultaneously.