Server Den: Cisco Turning UCS Into Server Battleground

I've spent the past month poring over survey data and working on my State of Server Technology 2010 research report for InformationWeek Analytics. (I want to be a tech analyst when I grow up.) So my interest was piqued recently when Hitachi Data Systems said it was getting intothe Unified Computing System space.

This is an interesting development mostly because of what it says about vendors' marketing strategies, and their perceptions of where enterprise buyers' heads are at. I've long assumed that executives at the big server guns -- IBM, HP, and Dell -- must be blanching at the runaway freight train which the UCS moniker has become. (Oracle/Sun, perhaps less so, but I'm waiting until a pending interview with them takes place before I write about their strategy.) I can image those guys figuratively screaming, 'Hey, we've got this stuff, too, and we were here first."

One reason Cisco is gaining ground in the compute portion of the data center so rapidly is because UCS is a more accessible term than what the other guys use. (Note that this doesn't have anything to do with the intrinsic quality of anyone's server, storage, and networking resources, which are top notch all around. I'm talking about the marketing battle.)

Dell calls its best-of-breed data-center approach the "Efficient Enterprise." HP goes with "Converged Infrastructure." IBM uses "Dynamic Infrastructure"--terminology that's about to take a back seat to Smarter Systems for a Smarter Planet, which is favored by IBM senior vice president Rod Adkins, who's the new guy in charge (see see my interview).

What I found in researching my State of Server Technology 2010 report is that UCS, while nominally a data-center play, has also positioned Cisco squarely in the server market. Enterprise buyers are, increasingly, interested in taking a look at the blade and rack servers bundled into UCS.

This makes perfect sense, since taxonomically servers are a notch down from the data-center bucket. It also makes sense that Cisco's UCS starts to hit its sweet spot in the middle-to-upper-level of the midmarket, because that's where you start to find enterprise customers with the bucks to buy full-blown data-center configurations.

That's as opposed to small, small businesses, where there's more of a tendency to jgo out and buy a Dell PowerEdge or HP ProLiant server, and plug it in. Very small companies have neither the bucks, expertise, nor often the need to muck about with big data-center architectural visions.

In that mid-and-upper mid-market spot, you have that need, and some money, yet perhaps not the same wealth of knowledge present in a large enterprise. That's why the one-stop-shop presented by UCS can be so appealing.

Hitachi's entry, under the UCS banner, says they see this, too. IBM's movement to what they believe is the more accessible "Smarter Systems" branding is additional confirmation. Personally, I believe HP may well air out its marketing terminology at some point, too. Dell, with its already differentiated "Efficient Enterprise," will probably stand pat.

For its part, Cisco might not be too happy with what I'm written above, which seemingly dumbs down what they've accomplished into a "data center for idiots" approach. Honestly, that's not my intention or my perception, though it is an easy, reductionist way to present its positioning in the market.

To dive deeper into some of the nuances surrounding UCS, I spoke recently with John McCool, senior vice president and general manager of Cisco's Data Center, Switching, and Services Group. He provided some interesting perspective. I was particularly interested in the insight on how UCS enables easier management of virtualized resources. We also touched on how network resources are going to be sucked up more by video bandwidth requirements, and what that means. Here's an excerpt from our talk:

InformationWeek: What's the big challenge?

John McCool: A data center architecture, which can host multiple applications and can scale up or down across the infrastructure. The unified compute, unified infrastructure, and unified fabric approach we're taking allows that first step.

As companies look at that they say, OK, now I can move virtual machines and virtual workloads across a vast array of x86 resource. How am I going to move that from a set of machines [to] across the network or maybe across infrastructure as a service, across the cloud? That's when they understand this is really a networking problem.

We've introduced new technologies like Cisco Overlay Transport Virtualization (OTV), which allows virtual machines to move across multiple data centers. So you can start to envision moving your workload between two sites using that technology for disaster recovery, or in the future being able to move those workloads to a service provider that's providing hosted services.

InformationWeek: What are the pain points?

McCool: We're seeing customers who have done a lot of virtualization. Now they're struggling with how to get control over that virtualized environment. Then they want to know how to take advantage of it terms of future movement and further optimization. They've reached kind of a plateau.

InformationWeek: Let's look at the network future. There's going to be a step-function increase in demand because of video. This will require smarter software, more analysis of packets, and more decision making. It's like a new take on the old cliche; it's not your father's router anymore, correct?

McCool: You're absolutely right. I think what you're trying to capture here is, it's not just about how fast things move through the pipe. It's the amount of processing and intelligence on top of that. What's happening is, additional technologies are converging on these networks--security, wireless integration, policy. This does require more processing; either more traditional compute-type processing on packets or hardware-based processing.InformationWeek: What's going to be the focus for the next 12 to 18 months?

McCool: The borderless network, and how users are connecting to the network. Increasingly, there are more non-traditional-IT devices -- industrial devices, smart-grid devices -- that need to connect back to the data center. There's also the changing role of users and globalization, which are impacting the way enterprises are deploying their networks. That's a huge area of focus for us.

InformationWeek: Does this mean that the complexity of consumer networks is going to soon outstrip that of enterprises?

McCool: The consumers accessing the network through the consumer and service-provider network increases the complexity of the enterprise problem. A decade ago, we were probably all issued computers in our office or a laptop from our company. That method of controlling user access to the IT infrastructure is losing ground. We're expecting to access any information anywhere, anytime, including our corporate information from our smartphone, from our home computer, from a computer at a hotel. That requires IT to look at how they protect their assets.

Now, as a user and a consumer, I'm expecting to be connected all the time. That means I should be able to connect to my company as easily as I connect to Facebook. This should be a seamless operation, but still be secure from an enterprise standpoint.

InformationWeek: It sounds like this raises security, policy, and capacity planning issues, all at the same time.

McCool: You got it. That's it.

InformationWeek: I'm trying to understand what I perceive to be the dual nature of Cisco. It's an enterprise networking company, but there's also telepresence and consumer-oriented stuff like the Flip video camera.

McCool: We look at ourselves at our very core as a networking company. What we see converging on top of those networks are three things: Video, the interconnection of the data center on top of that network, and the collaboration stack that then rides on top, which we're using today through things like telepresence.

InformationWeek: Will video will be the humungous capacity sucking application?

McCool: It changes the nature of the way networks need to be built, both at services providers but also at the enterprise and in homes. If you think about our consumer thrust and what we're doing with Flip, that's about consumers driving video and more of a seamless operation in how they deal with video over the Internet.

InformationWeek: It also elevates QoS considerations, correct?

McCool: Absolutely.

Recommended Reading: