While Cisco Systems branches out into consumer electronics, video, mobile data and other areas, one of its biggest areas of focus today is enterprise data centers. The dominant LAN provider thinks the transformation of data centres through virtualization calls for new kinds of connections and a broader role for network technology and intelligence. John McCool oversees this push as senior vice president and general manager of Cisco's Data Center, Switching and Services group. We talked with McCool recently about the company's own role and its partnerships with other vendors in this arena.
Can you give me an update on Cisco's relationship with EMC and VMware around Vblox?
McCool: Let me step back a little bit and give the general context. As we start to see a lot of changes from a technology perspective and our industry moves to cloud computing — how people are consuming technology, where they're getting information, which data center — the need to converge a lot of these technologies is getting increasingly important. And whether we build ourself, buy technology, or partner, the ability to execute in all three of those dimensions is critical.
Vblox is one component of several areas that we've partnered with both VMware and EMC. The initial partnership with VMware, I would say, was born out of customer demand for making virtualization work and scale in data centers. They saw the promise, they were using it effectively to lower their costs, but we started to engage with VMware in areas where it was difficult to scale across the network. The network designs wouldn't let virtualization move in a large-scale way. And we made a series of investments in technologies that allowed broad-based scaling within racks of machines and across racks, and now moving from data center to data center. Vblox is a piece of the work that we've done in that context. It's about making it easier to consume and service the combination of storage, virtualization, and compute — and network.
So, we're in the market with Vblox and pleased with the progress and moving that along. It came on the heels of our Unified Computing introduction, (where we are) quite pleased with the progress. We announced some additional customers like Taser (the maker of electronic stun devices), who's using it for a video-based application, which we think is pretty interesting.
It's been almost a year since the announcement of UCS (Unified Computing System). What more can you tell me about the progress on the UCS front?
McCool: What's interesting in (Taser) and several examples like that is its adoption … a middle market that we hadn't quite anticipated when we came out with the products. Companies like Taser see it as an opportunity to consolidate lots of smaller servers, which traditionally they had been buying, and replace those physical servers that are disparate and bring them together on a physical machine. Initially, we compared (UCS) to the large data-center opportunity, but it's having a tremendous value proposition in that midmarket, which never really went through this phase of consolidation. They're skipping the physical consolidation and then going right to virtual consolidation.
The other component is its relevance in service-provider data centers, where they're interested in hosting their own services to add value to their customer base or their customers' data centers. The multitenancy of the architecture has been really appealing.
And then the third component, I'd say, is that there is a class of enterprise customers who look at deploying the network-based technology and the UCS together to offer services within their enterprise. So they may be ones that had multiple data centers in different divisions or different functions and look at this as a way to emulate those multiple data centers, but consolidate them virtually. There have been more of those, also, than we suspected.
In some segments of large enterprises, ones that may be divisionalized or have different segments of their business, (where) each segment may have run a separate data centre, they see the opportunity to bring that back into a single data center but still provide the consistency of services they had in the multiple data centers.
So, a division wants some control over its servers and they can continue to have that because it's virtually their own data centre?
Is the partnership with EMC all that Cisco is going to need in terms of storage?
McCool: We'll respond to whatever changes in the marketplace transpire around storage or virtualization or networking, either through our own internal development or continuing in a partner model. Right now we feel extremely well-positioned through the partnership and also a broader set of partners on how we can bring this to bear around the network architecture. We'll be very focused on the nature of storage traffic and how it moves across an IP-based network, and what that means to the cloud. Look at our core business and what has to converge to make that happen. Obviously, we've made the first moves by bringing together Fibre Channel over IP, which reduces the network complexity in operating over a storage environment. The fact that now you can run any compute system, either with network-attached storage or block-based storage, without having to worry about pre-provisioning what network type, has inherent flexibility. It's an important component as you start to move to more cloud-based architectures or data centre-to-data centre architectures.
You have to take the view of how we look at the entire problem. We'll always look at the challenge around the data center, or this compute-storage environment, from our core of switching and routing. That's the way we approached the opportunities we saw with VMware and EMC. Where can the core network provide value to those systems? If we looked at computing, we saw that there's been a lot of innovation at the chip level, a lot of innovation at the operating system level, and the real next step in those computing platforms was how they integrated with the network. That market transition around virtualization really let us take network technologies and drive them in.
We'll take the same approach with storage. Are there opportunities for us to build around the network to make it better for our customers to run their storage environments?
Today we feel that we understand where this is going from a storage perspective, but it's a pretty dynamic market with this transition around cloud and virtualization. If there's other opportunities where we can add value to the storage environment through the network, that's how we'll approach it.
Two of your largest competitors, HP and 3Com, are coming together. How does Cisco see that, and what do you think it's going to do to the market and your competitive position?
McCool: You're seeing a lot of moves in the marketplace based around the network. A lot of folks are focused on the company issues and who's getting in. But I think thematically, the issue is that everyone's saying that the network's going to be extremely relevant to the infrastructure challenge in enterprises and service providers. In a data center, the network continues to be about 15 percent of the power consumption and 15 percent of the spend. People are starting to get that, as workloads start to move across the network. Unless you understand that technology, and unless you can make it more uniform, you can't bring an important piece of the value proposition to the customer. So I'd set the whole thing in that context as opposed to this intergalactic, large-company kind of thing. There's an appreciation here that the network is more than just a single switch or a single component, that there's a richer portfolio — not only portfolio of product, but portfolio of capability to solve a networking issue.
What steps is Cisco considering to reduce power consumption in data centres?
McCool: If you look back three or four years ago, people were focused on the power consumption of an individual thing — a part, a product, a phone. People are starting to look at it from a more systems-oriented perspective, and often you can see that actually increasing the power of an individual device lets you reduce the power of the overall data center. You can look at higher-density and higher-capacity network equipment, but use less devices as a result. You can take multiple adapters, multiple connections to the network that may be lower power than a new adapter, but converge multiple of them on a single adapter. This is kind of the approach we're taking, is to let more happen with less devices that may individually have a little bit more power but radically change the footprint.
The UCS system is higher powered than a single server in the large configuration, but its power consumption per virtual machine radically changes the profile. So you can reduce the number of boxes, number of connections, number of I/O, by moving multiple of these server systems down to a UCS. We've seen that even in our own designs that we've done around Nexus and UCS in our data center in Richardson (Texas). We reduced something like 4,800 cables alone.
In the past, especially before the downturn, people were building out new data centres for capacity. Huge expense in the physical shell. Hundreds of millions of dollars to build out that real estate. And I used to ask customers, when do they retrofit? And they'd say, well, we don't have time. The focus now has shifted to, “How can I better utilize the existing footprint that I have?” Our data center was a great example of that. We needed more capacity. So our design point in our data center was increasing capacity for the same physical shell, the same power footprint. We went to virtual machines to do that, so there was a dramatic increase in the number of virtual machines.