We would be happy to give you a full briefing on the RamSan-500 enterprise flash solid state disk and its performance characteristics. As you can see above, we provided some data to compare the RamSan to the STEC drive but you can only take so much from that in terms of comparing us to the EMC system. EMC has not published any real performance data so it is really hard to judge how well they are taking advantage of the STEC claimed performance. If you have looked at the EMC controller-cache-storage architecture you can see that this architecture will add latency to STEC drive accesses. You can also see that the EMC backend loops have an interesting effect on the actual deployment of the drives. So, all told we are left where we usually are with EMC and that is that they do not publish real performance data and we will have to determine their actual performance by tests that real world customers do comparing our products.

Woody Hutsell, EVP, Texas Memory Systems, 713-266-3200 x.241

The 30x claim by EMC made me smile too. While I have proved through physical bechmarking of the STEC drives, they can sustain the claimed 50K and 19K figures when using 4K transfers. As you increase the transfer size, the IO/s decrease accordingly. BTW, the only controller I could use to actually drive these was SVC.

So the 30x statement, I read, to mean thats all that DMX can actually get from them. This more a statement of the FCAL capabilities at the backend of DMX than anything else. I guess its also maybe a statement about their pricing being 30x. I know the price STEC were talking when they talked with us, and also that EMC signed an exclusive deal with them for the FC variants of these drives.

These drives are 2Gig FCAL attach, so their 4Gig backend will be slowed down when you put these in, not really a problem for IO/s but will limit the MB/s claims.

What I find most interesting, and probably again the reason for the 30x (rather than the 150+x the drives are capable of) are that they are supporting up to 32 per DMX Quadrant. Now, we know that EMC don’t publicly benchmark their controllers, IBM does and so we have an idea of what a DS8300 can sustain. The DMX is likely to be in a similar ballpark, based on the number of drives they supprot etc etc. So don’t be expecting the drive performance from the array, its going to be limited by the physical subsystem and not the drives. This is a capacity play, and fair game. Its a reduce latency and much increased IO/s range, even if you can’t actually get the max out of the drive.

The drives themselves meet the claims made by STEC. They are sustainable in the right enclosure

I’m very interested in the performance claims here, since EMC has traditionally been quite reluctant to make any performance claims about their products. But I guess when you introduce a super-expensive technology just for a performance boost, you kind of have to talk performance numbers!

I bet the STEC numbers are “best case” and the EMC numbers are “safe” – this would reflect the marketing position of EMC and explain the discrepancy. I’d love more information the TMS configuration…

Stephen

Maybe there are some issues with the drive in situ.

Enjoy the wheels