Enterprise Flash Drives and DMX -- The Aftermath

Every now and then, I get a ringside seat on a textbook example of human behavior in this industry.

And, I have to share with you, the reaction to EMC's recent announcement of enterprise flash drives for the DMX was very interesting, less about the factual content of what was said, but the different positions everyone took, and why.

And, going back over past experiences, I'm realizing that the last few days is a template of how a fiercely competitive, vibrant industry (e.g. storage) reacts to something truly new.

What Happened

On Monday, EMC announced a game-changer, from my perspective.

Here was a cool technology (enterprise flash drives) that delivered mind-bending performance, packaged in such a way that they could be put immediately to use, enveloped by a mature architecture of functionality, and well-supported by the de-facto industry leader.

I think many people expected this to eventually happen, just not so quickly.  And certainly, not from EMC -- they were expecting it from some small startup, perhaps.

The different reactions were fascinating.

Wow, It's Expensive!

I don't know why anyone was surprised by this -- memory is more expensive than disk, right?  Given that enterprise flash drives offer the same transactional performance as 30 traditional drives, and respond 10x faster, you might expect that they'd be a bit pricey, right?

I think three key points might have been overlooked.

First, they're not for everyone.  We know more than a few customers who can point at a particular application, and make a direct correlation between improved performance and improved revenue.  For these folks, they can't wait to get their hands on these puppies.

Sure, that's not usually the case for garden-variety applications like Exchange, but can be case for something like a currency trading application. 

Second, no one is gonna fill an array full of these babies anytime soon.  In a tiered environment, think in terms of a few 146GB enterprise flash drives to support the information that absolutely needs this level of performance.

You probably won't need a lot of them to rock your world, performance-wise.

Third, remember flash memory follows Moore's Law, not Seagate's Law.  Prices should decline much faster than disk drives, if history is any indicator.  If you track how DRAM costs have been declining over the years, that's probably a pretty good model.

Anyone Could Have Done This

There was a school of thought that this little exercise involved nothing more than a trip down to the local parts store for EMC.  Buy a standard widget, and just plug it in.  What's the big deal here?

Far from it.  For the complete story of just how much effort went into this, I'll refer you to The Storage Anarchist's definitive post on the subject -- excellent reading if you're interested in gory details and a fascinating story.

It's Been Done Before

A number of industry watchers strangley offered up this perspective, which I found surprising.  Maybe they didn't have enough time to really understand what we had done before they were asked for an opinion.

Sure, I'll agree that memory-used-as-disk has been around in the industry for a long time (EMC used to make an SSD device -- Orion -- way back in the early 1990s), but they missed a few key points that I thought pointed to the uniqueness of the approach.

I guess everyone has their own perspective on what's truly unique and what's old hat, but as we share what we've done with customers, they see it as unique and new, and I guess that's what matters, doesn't it?

The Typical Competitive Reaction

If you're in this industry, and a competitor does something really cool, you usually have only two public responses: (1) it doesn't work, and (2) nobody wants it anyway.

Meanwhile, you call emergency meetings with your engineering leads to figure out just how quickly you can start telling customers "me too" in your roadmap.  Good luck, guys, it's more work than you might think.

Not surprisingly, competitors reacted publicly as predicted. 

Many pointed to the wear factors associated with normal uses of flash memory.  C'mon, guys, we're not idiots -- we wouldn't ship anything that we hadn't tested the bejeezus out of and hadn't convinced ourselves that this would stand up to some of the most demanding applications on the planet.

And, of course, there were a few who sniffed "we haven't seen any interest in this technology".  If that's true, I might conclude that you don't have any customers who use your products for high-performance applications. 

I'm sure you didn't mean to imply that.

And The Truly Strange

I've got to hand it to Tony Pearson who blogs for IBM's storage group.  If there was a prize for creative competitive responses, I think he deserves special recognition.

He managed to come up with a truly unique theory that has to be fully read to be appreciated.  BTW, I read Tony regularly, simply for the entertainment value (?) of some of his posts.  Like this and this and this.

The Bottom Line

Look, I haven't met a customer yet who doesn't want to use enterprise flash drives, all things being equal.  As of today, they're new, and they're expensive.  But they work as advertised, and they deliver an amazing advantage to people who really need that level of performance.

But as with every Big Technology Shift, someone had to get the party going, right?  Someone had to put a product out there, make it work well, and begin the gradual shift to enterprise flash drives as primary storage, and eventually relegating traditional disks to lower tiers of performance.

This won't happen overnight, but -- make no mistake -- it will happen.  Sooner than some people might think.

I am heartened that more than a few people out there took this future-looking perspective, people who aren't affiliated with EMC, and offering that with this announcement, the way we think about enterprise storage would never quite be the same again. 

Expect More Rock Throwing

Part of me knows we're going to see an intense war of words around this topic in our incredibly competitive industry. 

Why?  In our little corner of the industry, this is really big stuff, and most everyone knows it -- even though they might not want to admit it.

Not surprisingly, there are a few pundits out there who don't particularly like EMC, and absolutely hate it when we do something right.  Some seem to make a career out of throwing rocks at us.

And the competitors?  I think their strident tone will change a bit when they eventually figure out how to do something roughly similar, or -- at least -- announce their intent.

Until then, sit back, reach for the popcorn, and enjoy the show!

I sure am ...

Comments

[Tim]

Would that be sort of like the symmetrix and thin provisioning? Excuse me "virtual provisioning"?

[Shibin Zhang]

This is another wonderful article.

I have the same opinions (Believe me: not after reading this article. I had never been an architect, but I never stop learning and thinking). However, I would not just enjoy the show, I would think deeper on the new opportunities.

Shibin