The Coming Architectural Wars

Much has been made online of Cisco's rumored plans to enter new data center markets.

It's easy to dismiss all of this as just one large enterprise IT vendor poaching on another's turf.

But, to many of us, there's much more here than initially meets the eye ...

The Facts

Now, I can't comment directly on all of this -- being bound up in a welter of various NDAs -- but there's plenty of interesting reading on the internet here, here, and here for starters.

Of course, this is all unsubstantiated rumor until we get an official announcement from Cisco, but -- assuming there's some germ of truth here, what's going on?

I think it's a fundamental shift in power in how enterprise data centers will be built.

It All Starts With Virtualization

For the last few years, I've been routinely stating that "virtualization changes everything".  And it jsut keeps on going, doesn't it?

In a fully virtualized environment (think VMware at scale), application and information containers move back and forth, blissfully ignorant of which server they're running on.

If you assume that this style of computing becomes the dominant model, you start thinking of servers -- and optimal server design quite differently, don't you?

As a simple example, since VMware basically pools all resources, wouldn't you like a server and interconnect architecture that basically does the same?

There's a big difference between a server that can run in a virtualized environment, and an architecture that's intended to do so at scale.

But Management Changes As Well, Doesn't It?

For those of you who have realized that legacy management approaches struggle in a virtualized environment, virtualization at scale forces an entirely new way of thinking.

You'd like to be able to provision, manage, orchestrate, secure, etc. multiple virtualized entities in a single motion.  And remember, these entities aren't tied to a physical device or location anymore, right?

The traditional server-oriented model doesn't stand up well in this light, does it?  Nor does the storage centric model, for that matter.

No, what's logical is to extend a management model that's already doing most of the job today, and teach it a few new tricks.  Hence the attractiveness of a network-centric management (and implied architectural) paradigm.

In This World, Does Everything Become A Network Service?

That's basically what's on the table here -- a new twist on the old concept of "network computing".  We just extend the model to include virtualized applications (server and desktop) as well as the information resource they need.

Layer in the discussion around private clouds, and it can be a pretty persuasive argument that *everything* gets viewed as a converged network service.

Scott McNealy is turning out to be either famously prescient or dumb lucky when he stated "the network is the computer".

The Predictable Reactions Have Started

Not surprisingly, entrenched data center server players like HP and IBM (and Sun to a certain extent) won't take this shift in architectural control laying down.

Indeed, the first few skirmishes have started (and have been replied to), but we're just in the early days.  Note: if any of you server bloggers would like lessons in Deathmatch Competitive Blogging, I'm sure that a few of us in the storage biz could offer a few lessons :-)

On one side, there's billions of R+D in non-x64 processor architectures (think Itanium, SPARC and Power), as well as much more in their proprietary flavors of UNIX, management suites, and much more.

On the other side, there's an entirely new way of doing enterprise-scale computing that doesn't necessarily value these legacy investments.

Should be interesting, no?

Where Is EMC In All Of This?

Well, if you're an outsider looking at EMC, you've probably realized that this sort of architectural thinking could align nicely with our own. 

Besides, it's not like we're exactly in love with IBM, HP and Sun, are we?

At a very basic level, people are going to need storage, back it up, tier it, etc. regardless of who wins the server architecture wars.  EMC has always been agnostic to server architectures, operating systems, etc. -- that's not going to change.

But, going further, you've probably seen how we've accelerated key data center technologies like FCoE, and done everything we can to make VMware work at scale.

And, it's not to hard to see how EMC's network-centric investments in resource management (Smarts) and information-centric security (RSA) can add even more to the discussion.

What Happens Next?

Probably one of the most epic battles we've seen in the server world for quite some time. 

But here's the good news -- customers will have entirely new choices that they didn't have before, and -- regardless of outcome -- that's always a good thing all around.

Comments

[John Webster]

Chuck; Agreed. Thinking here however is that Cisco will need more infrastructure stuff to really make Data Center 3.0 and beyond (i.e Data Center 4.0, 5.0…) a serious effort that could actually succeed, as opposed to one that just makes IBM, HP, and Sun grind their teeth. Cisco will need server and storage virtualization, systems management apps, and security for sure, and would also profit from content management software and, ah yes, storage.

We hear that Mr. Chambers is making Data Center 3.0 sales calls at the CEO/CIO level. I for one would bet that he’s very effective during these sessions. The question is: will the operations people that buy servers, storage, etc. go along with the Data Center 3.0 vision? I doubt that the network folks will be buying servers any time soon. So I think Mr. Chambers understands that to sell this vision and make it stick, he has to call at the highest levels of the enterprise and hope the Data Center 3.0 fairy dust trickles down so to speak. We think he bolsters his case when he can also weave Cisco-branded servers, storage, systems management, and Virtualization (that's big-V virtualization), into the fabric of the vision.

So clearly, EMC is a prime partner. You have everything else they need to make Data Center 3.0 a Cisco reality. But, what makes EMC a great Cisco partner, also makes EMC an interesting acquisition for Cisco. Hence, the embers of that old Cisco-buys-EMC rumor occasionally glow white hot as in now maybe?.

[Chris]

As someone who has been involved with x86 servers for a long time I welcome what Cisco is doing.

None of the other vendors mentioned (and there's another one based in TX you didn't mention) have been able to grasp the idea that the is going to simply be a box with compute power. They are still stuck in a 1990s mindset of "this widget makes my server better than yours." The playing field has changed dramatically in the last few years thanks to what VMWare has done and Cisco's rumored decision is a logical next step for them and anyone in the virtualization space.

[Chuck Hollis]

You're spot on as always, John ...

It's going to be interesting to see how Cisco constructs their ecosystem of partners as they move into this space, isn't it?

-- Chuck

[Chuck Hollis]

Chris -- agreed!

Every industry needs a good disruption every so often, and this might be the server vendors' turn.

We'll see, won't we?

-- Chuck