I ran many similar IOMeter tests when I first implemented my compellent storage array back in January '07. I was concerned initially with being able to generate the IOPS for the number of drives in my drive pools, so I disabled read and write cache for that testing. The results were that I was able to generate industry standard (100 - 150 IOPS) per drive in my drive pool. I would like to run your tests, but my storage arrays are now in production. The virtualization of drives does work very well and it tunes itself unlike EMC/HDS/IBM etc.

My initial system Compellent was 48TB configured for my SAN fabric of 125 FC hosts. We selected Compellent to replace a EMC Clariion CX600 and two HDS 9960's totaling 29TB.

There is no comparison regarding system management - Compellent wins hands down!

On the performance: we spent three months testing Compellent using JetStress and LoadGen. During the pretest build phase we generated over 600MBps of throughput against the Compellent front-end fiber ports and over 700MBps on the back-end for hours and hours at a time. The Compellent system just hummed along. The tests passed at over double our recommended loads for 6500 Exchange users with 7ms response times on the transaction logs.

When I moved my data from the CX600 (close to 14TB allocated) it only used 6TB of real space with Dynamic Capacity. This solved a huge SAN problem for me with EMC and Hitachi 9960's.

I am just budgeting to implement VMware this fall, but I have high confidence that Compellent is the right SAN storage solution.

-dave