Interop: 5 Questions For Cisco's Jayshree Ullal

For the past year, Cisco Systems has been hammering home its Data Center 3.0 vision. The San Jose, Calif.-based vendor has released a host of products that for the data center, from the new line of Nexus data center switches to a new fibre channel over Ethernet offering.

Shortly before her keynote presentation at Interop Las Vegas 2008, ChannelWeb caught up with Jayshree Ullal, Cisco's senior vice president of data center, switching and services. Ullal talked about how customers and solution providers can prepare data centers for the new influx of sophisticated applications, like video, and discussed how over the past couple of years Cisco has "gone from wanna be to we're here to be" in the data center.

Cisco has made several key announcements in the data center lately and the increase of applications; what do you plan to discuss this year at Interop?

I'm going to talk about how the enterprise is transforming and one of the transformation aspects of the enterprise is we're moving from siloed applications on the network to increasingly understanding how new applications like video and Web 2.0 can interact not just with the network in a loosely coupled fashion but really tightly with it. That's a key piece and video is an example of that. Much like we did with unified communications where we really focused on the right power over Ethernet and quality of service required to run it, video requires us to do a number of things: How to look deeper into packets to understand the type of video traffic; how to enable end-to-end predictable latency. Video traffic consumes lots of bandwidth it can be anywhere from a few megabits to 20 megabits, but making sure you can send that information reliably so when you see it visually there's no jitter is very fundamental.

As video and these bandwidth intensive, sophisticated applications are moving into the data center, what do customers need to do to get ready for this?

Three things: First you have to future proof the network to have the right bandwidth capacity, performance, etc. Performance is not just bandwidth and capacity, but latency. The video streams need to be appropriately prioritized. You may have a 10 gigabit pipe, but if you're sending a large database application and you're not really putting the focus on video then you really haven't appropriately created the right virtual main for video. The right level of pipes, latency and jitter is important. That's one.

The second is recognizing the video application. What we've done in Cisco and a lot of our switches, the Catalyst 6500 is a good example, is built the right packet inspection capabilities so we can identify video streams. When you have this kind of network based application recognition you can say ok here's my video stream and identify and give it the best choice of path across the switch. It's not just providing the bandwidth but providing the right path for that bandwidth.

The third thing is, not all video is the same. You can have very high-end TelePresence video, you can have desktop video and digital surveillance. The ability to adapt your video streams from being static to interactive to highly scalable means you have to have a platform that can deal with few flows to millions of flows, especially when you look at a large service provider network running across enterprises. This is becoming more and more important because when you look at enterprises today it's not one size fits all. More and more the enterprises really are about software as a service and the ability to run multiple applications including video.

These Web 2.0 applications allow you to not just accelerate packets and protocols but in many cases accelerate the network sessions as well. You can look deep enough into the packet and look at if it is an HTTP session or a URL session and accelerate it that way. That's the beauty of what you're seeing with the new generation of enterprises. Not only are you moving bits fast, but you're actually moving the application associated bits faster. So you have to distinguish between a dedicated video stream and also the reliable data applications you want to run and make sure you can do these types of things.

NEXT: What's Data Center 3.0 Mean For The Channel?

With Cisco's data center strategy and the applications built into it, what does this do for channel partners?

Much like unified communications, data center is really a practice it's not just a product. We now have a whole portfolio of products, everything from Nexus unified fabric, Catalyst switches, storage MDS switches, high-performance compute with either infiniband or Ethernet low latency switches, security with TrustSec to bring in the data center encryption and firewalling capabilities and all the management and orchestration pieces.

What our channel partners are now equipped to do is: We have the vision, the data center 3.0 vision with consolidation, virtualization and orchestration. We have a rich product pipeline What we need is close partnership with the partners to really deploy this portfolio of products and build a data center practice. There are data centers the way they are constructed today which is much more inside and there are next-generation data centers where the network is truly the platform for orchestrating the compute, storage, all the different machines and applications. Our partners are a very vital piece of this. We are kind of the fabric. Our partners can make our fabric come alive by connecting it to the right storage machines, the server machines and the applications and becoming the systems integrators.

What do you see as the challenges the channel and customers face with this data center strategy?

A lot of it is education, training and practical deployment scenarios. The vision is very high, most of us get graded a nine out of 10 or a 10 out of 10 at Cisco for having the vision. The product pipeline with all of the new product introductions we've made in the last year is equally high. The biggest challenges lie in the deployment scenarios. How many of these things can you do in a greenfield data center versus a legacy data center that you may not have enough freedom of rotation, you don't have the real estate, you don't have the power and you don't have the budget. There are at least four or five aspects to a data center consolidation, virtualization, security, application delivery and high availability; in a greenfield data center you may be able to tackle all of them. In an existing data center you may only be able to pick one of them and do it right. Making those trade-offs are what's the hardest issue in an existing versus a new data center and how do I go tackle and insert in a situation where there's a lot of moving parts already is probably the greatest challenge and opportunity.

When partners and customers are going to be deploying these next-generation data centers, what's the one thing they need to know when evaluating their data center strategy? There's a lot of new technology there and a lot of new applications, what do they need to know?

There's a physical aspect of how large is the data center and the power and cooling. There's a facilities aspect comes in first. The second is, what are they trying to service, what is the level of compute capacity and storage capacity and once they can profile the class of machines and the total amount of capacity then they'll have a much better idea of how much of a foundation of a network do they build. It's much like constructing a house. If you know how many floors you have in the house, which is equivalent to the storage and computing capacity, then you can go and decide what the foundation is. Then the applications, which are more like the faucets, the bells and the drapes and things like that are equally important today more than ever. Before we just put the infrastructure in and said the applications come later, today you have to pay a lot more attention to are these are database applications, are they Web 2.0 applications, are they video applications, are they file applications, are they legacy mainframe applications and treat the appropriately. The beauty is many of these applications work much more tightly with the network than they used to and don't require a lot of middle ware or specific consultants of the past. Consolidation both of physical facilities and the network, virtualization, who needs what and what capacity in the network, in storage, computing; and automation of the applications are the three hot buttons they really have to pay attention to.