StorageMojo’s rigorously unscientific survey
My absolute favorite product – and I haven’t the faintest idea whether it works or not – is the StoreWiz compression appliance. It sits in front of your big NAS box – the more expensive the better – and doubles or even, they claim, quadruples your NAS capacity through in-line, wire-speed, data compression. Mindful of the whining by /.’ers after my post on Diligent’s 25x backup compression I asked them how this was possible. The recent U.S. hire – the company is based in Israel – said they use Lempel-Ziv compression.
Now I do know something about LZ compression from marketing the first DLT drives, and 4x compression is not in the cards for normal data. I said that and the guy said, “It’s magic.” Finally, a simple, clear message from a storage company. That did it for me. The old ways are best, and what is older than magic?
If anyone has tried it please let me know how it works. They claim 120 customers. They said the 100 MB/sec went for $22k and the 220 MB/sec model was $42k, as I recall. StoreWiz guys: you are welcome to respond.
Most stunning performance – and boring demo
I stopped by the Texas Memory Systems booth, to say hi to Jamon, the SE who wrote the TMS response to the StorageMojo article on flash SSDs and his boss. Nice guys.
We’re standing next to a couple of wide-screen LCD displays that have a circular dial on them, like a speedometer, and a caption reading 278,623 IOPS – something like that. I thought it was a slide, and then I noticed the number changed a little every so often. “So what’s this?” I said, pointing to the display. Jamon said they were running some benchmark (IOmeter?) and that they really were getting 278,000 IOPS 278,000 IOPS right now. As we spoke.
So I suggested that maybe the demo ought to move or something. You know, so it wouldn’t look like a slide. And they agreed. Oh, and here’s a new slogan for TMS:
“The performance of 2,000 disks at a tenth the price” © StorageMojo.com
Really, is that so hard? TMS, if you want to use this, call me. Everybody else, do you think this would better help your pointy-haired boss understand why RAM-SSDs are such a good deal?
Best freebie
Nimbus Data is giving away a free Windows 2003 iSCSI target. Now you can turn an old server into an iSCSI storage server. Works with Microsoft’s free iSCSI initiator. Even better, you don’t have to be at SNW.
Other stuff from SNW coming soon. Stay tuned.
Comments welcome of course. Please fill in the gaps with your selections.
Robin,
As you say, a ‘ rigorously unscientific survey’.
To comment on the 278,000 IOPS we need to know how ‘small’ the transfers were, how many host channels & servers were involved and if this test was a read, write or a % mixture of both.
Richard,
Good question that I wish I’d thought of myself. I’ve asked TMS to respond and I hope they do.
Robin
Hi Robin,
Thanks for pointing out this question.
For that test I we were doing 8k Random Reads with a high outstanding IO count (40 total), Six 4 Gbps links from 1 host, a dual cpu dual core Opteron with QLogic HBAs.
The site below references that they use a RamSan-400 for the test and shows close to the performance you can get with one of the RamSan dual ported FC400 (4 Gbps) Controller for a variety of loads. The RamSan-400 supports 4 FC400s (8 ports total) with active/active access from each port to any LUN in the RamSan.
http://www.qlogic.com/documents/datasheets/knowledge_data/whitepapers/Scaling_SAN_Fabric.pdf
The two limitations of the system (with maximum controllers) are 400,000 IOPS or 3 GB/s for any combination of reads/writes with varing block size.
Thanks!
Jamon Bowen
Sales Engineer
Texas Memory Systems, Inc