We tested several of the arrays
We tested Pure, IBM (older version), Netapp, and Kaminario
None of them were slouches though in our tests the Netapp seemed the slowest and most of the benchmarks indicated it was. Now this was the older version the new one may be better. It was using the Santricity software we we were familiar with from the e5400s and overall was pretty good. The Kaminario was the fastest in the widest range of scenarios but the amount of wiring due to their architecture was not something we were interested in but overall was most definitely the fastest in all scenarios. The older IBM array was limited by its 4 8Gb ports the new one I hear has more. Though it was fast in all the scenarios except in the full sequential throughput tests limited by the 4 ports. The management interface was lacking in our view.
The pure was not the fastest in several of the scenarios we test but it was fast. Its reliability was fantastic and upgrades/failovers were completely transparent. Though we did not buy it for our original use case testing we did buy several for our VDI/VMware environments and have been totally happy with them we can't stress it during our normal workloads and short of running iometer on a bunch of vms at once its tough to make this thing sweat. It is a bit slower on writes than many of the other arrays we looked at but good enough to make us happy. The management is dead simple though we would like maybe a few more metrics it is adequate. The dedupe and data reduction has been right in line with their quotes overall very happy.
None of them were bad products and all of them had advantages/quirks but we bought the best one for our need (granted not our original need We went local flash for our original needs) I am sure all of them will only continue to get better and kill most of the classical san arrays in many scenarios we took out several racks of netapp equipment and replaced it with one pure for vmware for example.