I don't know about you, but I love watching the back-and-forth in this industry.
Especially when something big and disruptive enters the discussion -- like flash being the long-term potential replacement for more and more uses of "rotating rust" -- that is, disk drives.
OK, So We've Got A Vested Opinion Here
You all probably know that EMC raised a small ruckus in the industry when we announced EFDs (enterprise flash drives) for the DMX back in January of this year. More recently, we made them available on the new CX4.
For 2008 (and perhaps a decent part of 2009), it now looks like we're the only game in town -- at least for the time being.
So, even though you should be taking my comments with the prerequisite grain of salt, it's still big fun.
How The Discussion Progressed
Now, this is not a clinical recounting, just my impressions.
The first round of noise came immediately after we did the initial launch at the beginning of the year. A small group of customers and analysts took the perspective that, well, this was big.
But -- numerically speaking -- a larger number of industry watchers took a more negative stance. Performance problems, they claimed. Wear-levelling problems, they claimed. No industry standards, they claimed (?)
No amount of factual response from us or anyone else seemed to sway them.
I can't say whether or not that was their independent perspective, or they were swayed by the rest of the storage industry that was caught a bit flat-footed, and needed some help making their case.
Regardless, there was an initial burst of extreme -- almost violent -- negativity and cynicism.
And then, quietly, we starting hearing from a variety of sources that, yes, many of our competitors had been telling their customers that they'd be doing something along these lines later on, "when the technology was ready" -- a statement which leaves considerable room for interpretation.
Many of us knew that it wasn't as easy as just plugging in a few FC-compatible flash drives into an existing array. There's a significant amount of work that needs to be done to make these devices actually deliver the intended benefit, and not cause additional problems in customer environments.
A interesting thread then seemed to start with the server vendors: they mounted an argument that, since flash was really another form of memory, it belonged on the server, and not in shared storage.
Interesting perspective, that one. All fun to watch.
And, then, a few months ago, the cynicism and skepticism started again. Some sort of panel sponsored by the "flash industry" offered a very negative perspective on the prospects of flash supplanting disk. Even a notable financial analyst got into the mix with a decidedly cool perspective.
And, of course, the press picked up on all of this, which reverberated around for a while. One could argue that a somewhat negative consensus was in the process of being formed -- thanks, in no small part, to the tireless efforts of many who weren't really ready for this industry transition.
A few internet curmudgeons even took a few swipes at our executives (Joe Tucci and Dave Donatelli) for our bold predictions around the impact of this technology, and what we saw happening in the not-to-distant future. Don't think for a moment that any of this impacted our commitment to the technology one bit. If anything, we have the courage of our beliefs.
Besides, we were busy working with customers who had real-life situations where this stuff could offer a big benefit, so we got on with the work at hand, rather than debating the issue. These things have a way of playing out.
And Then, A Big Player Stepped Up
I found Intel's recent announcement that -- yes! -- they were going to be a serious player in this space, not only in desktop, but in enterprise-class drives -- one of those potential tipping points in the whole flash-vs-disk debate.
Not only that, but quick endorsements followed from -- yes! -- server vendors such as HP and Sun who now endorsed the technology, and said they'd be using it at some point.
So much for flash being only a server technology ...
Sure, it was the Intel marketing machine at work, but one has to admire the level of commitment that Intel demonstrated (a fairly comprehensive offering), their positioning (value-added to the core technology) as well as securing public endorsements from notable players.
Now, if you're arguing the "flash isn't ready yet", you'll be debating against EMC, Intel, Sun and HP, to name a few. And, also quietly debating with quite a few other storage vendors who aren't quite so public about their specific intentions -- yet (e.g. IBM, NetApp, HDS)
It's Always Fun To Step Back And Watch The Action
Yes, I tend to get wrapped up in all of this, and sometimes get too close to things that I can't really appreciate the broader dance around me.
But, in this case, I think we've got an excellent textbook study around what happens when something truly disruptive in an industry happens -- and how different players react.
Now, about that FCoE thing ...
Whoa there, Cap'n. SSD drives have been preannounced but are not available on the CX4. Soon, but not yet.
The technology is very viable and brings in to question why anyone would spend money on 15K FC/SAS. Also, does this negate the majority of the business market for 22K RPM drives, if they ever get reliable enough for the enterprise?
I've yet to see some good arguments around server-based RAM vs storage-based SSDs. I don't think there's a clear answer either way and I've heard them both, but not one that I could subscribe to.
Posted by: mgbrit | August 20, 2008 at 05:41 PM
Hi mgbrit ---
I apologize regarding the pre-announce vs. shipping distinction. You may be right, but I know customers who have them, and sometimes I miss the precise status at any point in time.
Lots of moving pieces here at EMC.
Regarding the 22k drive question -- yes, interesting -- but my first thought when I heard about this was -- is this about as relevant as a faster tape drive?
I think we'll see flash both in servers and storage arrays, much like we see with disk drives today.
I did not accept the mooted premise that all uses of flash in servers were inherently great, and all uses of flash in storage arrays were inherently wrong, or that flash only made sense as a "cache" for a traditional disk drive.
And I'd think you might be inclined to agree with me.
Posted by: Chuck Hollis | August 20, 2008 at 06:24 PM
Hi Chuck,
Your depth of understanding on storage topics is most impressive but I don't think you're a mind reader.
You've made a couple of comments not only in this post but in your previous ones suggesting that you know what competitors like HP will be doing. These comments do nothing to help the credibility of you or your company. Specifically, the comments I'd call out and my response are:
1. From this post: "For 2008 (and perhaps a decent part of 2009), it now looks like we're the only game in town -- at least for the time being."
>> I'm not really sure what you're saying here (for 2008, for part of 2009 at least for the time being???) but since we don't have you under a non-disclosure agreement, I can't provide specifics other than to say your information is wrong. Since you aren't shipping solid state with the CX yet, sounds like you can't claim "for the time being".
2. From your June 4 post: "...just about every storage vendor has had to significantly re-prioritize their roadmap, and figure out what they're gonna do about EFDs." (Enterprise Flash Drives)
>> I've asked everyone I can find at HP who knows our roadmaps and I can't find anyone who has briefed you on our plans around solid state technology. I can also tell you that the comments that Dave D made about SSD replacing fibre channel in 2010 didn't have HP rushing to re-prioritize our roadmap.
If you or your readers are interested in HP's view of solid state technology (versus your Carnac The Magnificent imitation), they can watch a short video that talks about it at http://h30431.www3.hp.com/?fr_story=1d62307e12719ec592505e59cd84f10c3fe68d95&rf=bm
I'd like to see you stick to what you know instead of speaking for the rest of the storage vendors. Maybe you should try a Joe Friday imitation (just the facts) instead of Carnac.
Regards,
Calvin
Posted by: Calvin Zito | August 20, 2008 at 08:27 PM
Well, I'd stick to the facts, but HP really hasn't provided any, so I do have to admit I have to improvise a bit ...
And, of course, if you had briefed me, then I would have had to sign an NDA, and then I legally couldn't have written anything about what I knew, right?
No, I'm not going to fall into that trap! That'd take all the fun out of it.
May I be so bold as to make a gentle suggestion for my esteemed colleagues at HP regarding this and similar topics?
Let's face it, you've had quite a while to sort out your story and positioning on this topic. This stuff has been in the marketplace for a while, hasn't it?
During this time, I have to say, it looks like your company has had its challenges in assembling a consistent story between its server group, its storage group, and other HP spokespeople.
Maybe my mind-reading trick doesn't work so well when the subject is in an apparent state of confusion ...
I think the answer here is simple. If you'd like to clear things up, simply tell people what your plans might be, and there'd be no need to hide behind NDAs, et. al.
Now that you mention it, I'm kind of glad I *wasn't* briefed ...
Thanks for writing!
Posted by: Chuck Hollis | August 20, 2008 at 09:17 PM
Hey everyone, go follow the link supplied above -- you just *have* to see the HP corporate message on solid state technologies!!
It made my day! I haven't laughed this hard in a while!
If you're in the bowels of the industry like many of us are, this may be the most entertaining 4 minutes you'll have this week!
Oh, BTW, the last four minutes of the video seem to be dead air.
Is there a message here that I'm missing? :-)
Posted by: Chuck Hollis | August 20, 2008 at 09:29 PM
Chuck -
I'm sure by EMC's standards, the video is dead air. But we know that tape is dead by that same standard, right? ;-)
The point of the video that you missed is that HP is taking a systems view of Solid State Technology. It's not just an external array technology but a server and storage technology and EMC can only address part of that equation.
You'll have to stay tuned to get the details.
Thanks for stopping by our StorageWorks blog.
Posted by: Calvin Zito | August 21, 2008 at 01:17 AM
I have to admit, it's good solid FUD.
Just for fun, imagine if I substituted the phrase "disk drives" in your statement, e.g.:
"HP is taking a systems view of Disk Drive Technology. It's not just an external array technology but a server and storage technology and EMC can only address part of the equation"
Because that's how customers now are thinking of it: flash as a potential replacement for disk.
Best of luck!
Posted by: Chuck Hollis | August 21, 2008 at 07:15 AM
My point exactly. Customers with ProLiant servers can move their disk drives from the server into an MSA giving them an easy transition from DAS to SAN - just move the drives from the ProLiant into the MSA and they are up and running.
This is in fact how customers look at it - they want an easy way to move from DAS to SAN and they get that with HP.
All the best Chuck!
Posted by: Calvin Zito | August 21, 2008 at 01:59 PM
I have to admit I love reading your blog. I have been looking for positions with EMC in the Boston area here and I cam across your site. Great insight at the latest technology. I agree that flash is absolutely ready and the fact that Intel is backing it aside from these smaller names like M-tron and sub divisions; we could see some real potential in the server market. I would be curious to see how the numbers compare to current top of the line offerings along the measures of sustained read and write speeds and random access times. Either way, no moving parts is a huge plus and step in the right direction. Hopefully one day the hard drive won't be a bottleneck anymore.
Posted by: Stephen DiSchino | August 25, 2008 at 01:23 PM
I am curious on this as well, as I've been eyeing SLC-based flash for my home PC, although not at $1000 per drive :-)
Practically speaking, though, the I/O characteristics of an enterprise flash drive are rather dependent on the array they're used in. For example, the DMX has large cache and rather unique algorithms, so we tend to talk in terms of application-level results, rather than low-level metrics.
Glad someone is reading and enjoying ... thanks!
Posted by: Chuck Hollis | August 26, 2008 at 08:50 AM
OK, everyone, you've got to see the latest from IBM -- they took the SVC, stuffed it with a bunch of flash, and are crowing loudly about the results.
Of course, it's not really a product, you know ...
http://www.ibm.com/developerworks/blogs/page/InsideSystemStorage
Funny thing about IBM these days regarding their storage business: they don't announce products, but they seem to announce things that aren't really products.
Weirder and weirder, I tell you ...
Posted by: Chuck Hollis | August 29, 2008 at 01:21 AM
So how many DMX quadrants would be needed to sustain > 1M 4K IOPs, at a 70/30 mix, you'd need at least 100 STEC drives runing absolutely flat out... with a backend storage system that could also manage the same IOPs. Oh and keep it under 1ms response time, without using cache...
How much floor space would that take up? An entire data-center? or the 1.5 EIA racks we used...
As for announcements, all will become clear...
Worried are we? It seems like it - yet again we've got you all talking about us, rather than yourselves... just adds to the suspense :)
Posted by: Barry Whyte | August 29, 2008 at 05:08 PM