So finally, we get the NetApp SSD announcement and I almost fell off my chair with shock! Is that it? You can stick a TMS RAMSAN array behind a v-Series? I was expecting something a lot more than this from NetApp; I was expecting something really clever!
Arguably IBM have already done this with their Quicksilver science experiment and is there anything especially stopping them qualifying the RAMSAN behind an SVC? What's to stop NetApp qualifying Clariion or DMX with SSDs behind a vSeries. Apart from the fact that it would make Chuck and Barry laugh like drains!! (EMC might even be tempted to qualify it just for the sheer amusement factor).
I expected more, I expected something which was going to force EMC to raise the bar on their SSD implementation. Maybe there's a lot more to come and maybe I need to wait for the mythical OnTap 8 but it's not lighting my fire especially at the moment.
I think that some really clever things could be done with WAFL to use SSDs in an efficient manner.
This is a big year for storage, time to step-up to the plate guys!
EDITTED: I had Texas Instruments on the brain today!! I knew it was TMS, just went a bit mad! It was the under-whelmedness of it all!!
Agree – the PAM cards are interesting (why did we have to wait so long for them? and PCI slot stealing isn’t good) but TMI behind a v-series (whilst good & fast) is not exactly innovation…
Heck even Sun have managed to do something interesting with SSD & OpenZFS, so just how have NetApp let this slip past…
Hoping this is a minor point announcment ahead of the bigger things with ‘actual GX’ & RAM & SSD 😉
Dude! It’s not Texas Instruments! It’s Texas Memory Systems – two VERY different companies!
And the proof is in the pudding: Although it’s “just” a TMS 500 behind a Vfiler, it seems to work awfully well.
Personally, I’m much more interested in the PAM and the implication that it can make big cheap disks perform like small expensive ones. Now that’s progress!
The PAM cards make things better…but not that much better. We had the first ones in the UK, performance of our SATA improved but it certainly doesn’t turn them into FC performance.
Woody at TMS has been trying to get us to qualify RAMSAN behind SVC for year – look at the interop matrix they get for free…
As with all things SVC and “tier2+” vendors, if we get enough RPQs it will be supported… and I believe we have… but we won’t be dedicating a press release to it when it is!
Look around, they’ve all been caught flat footed by last year’s Flash announcement and while I’m sure they’ll dig up some SSDs somewhere and claim parity not one of them have anything big in the pipe until 2010 at the earliest.
And well the PAM card isn’t that interesting, it’s just unprotected read cache on a PCI card.
Have you seen the cost of PAM cards plus licensing ? The monoliths have been throwing huge cache at disk for a long time with ever decreasing returns.
Hi Martin,
Thanks for covering our announcement today. For those wanting further insight there has been plenty of commentary (from competitors of course and others as well) on my blog.
-Val Bercovici
Office of the CTO, NetApp
Vice-Chair, SNIA Solid State Storage Initiative
http://blogs.netapp.com/exposed/2009/02/solid-state-sto.html
“The PAM cards make things better…but not that much better. Performance of our SATA improved but it certainly doesn’t turn them into FC performance”
Not what Netapp are positioning
http://blogs.netapp.com/storage_nuts_n_bolts/2009/02/spec-sfs-2008-the-benefit-of-the-performance-acceleration-module.html
Nice revisionist history / selective quoting John.
The rest of the story is that in this test, we swapped the 112 FC drives for SATA, kept the PAM cards in place, and got almost the same throughput (40,011 ops/sec) and overall response time (2.75ms) as the baseline, but with 75% more capacity. Also, at a 27% lower cost than the original.
Val,
I know your under pressure at the moment, but what’s wrong with the quote ? I referenced comments made from a customer, above, then pointed to a Netapp blog with the performance numbers. To be honest it’s late and unless I’m missing something real obvious, it looks like your comments just confirmed my conclusion.
Hi John,
Apologies for the confusion. For some reason, I thought you were picking a quote from a NetApp whitepaper, not a customer.
As I summarized above, for many workloads PAM can help SATA drives replace FC, particularly for aggregate random IOps.
There are of course exceptions, and if latency is more precious than overall ops, PAM 1.0 does not always improve that.
But PAM 2.0 … 😉