Q & A With Packet Design's Judy Estrin

Even when Judy Estrin is jet-lagged from day trips to the East Coast (such is the life of a high-profile board member), it doesn't take much prodding to get her talking about networking in general, and IP-related topics in particular. Of course, anyone who would start a company called "Packet Design" is a packet-head for life, something easily ascertained with a quick glance at Estrin's estimable resume.

Networking Pipeline caught up with Estrin last week, and got her opinions on a wide range of matters networking, including a look at what's interesting for the short- and long-term (like the combination of management and security) and why the first inch of networking might be more important than the last mile.

Networking Pipeline: What is your view on the "state of the state" of networking and IT? Is this a real recovery?

Judy Estrin: I think people do finally have some money. I would be very cautious to think that we're on our way up.

I think it feels good because for the first time in a while there's any budget, as opposed to budgets being cut. But I think that the sense out there is that people are only spending money on what they have to, only spending money if there is a really clear short-term return on investment. So that makes it still tough, in terms of identifying areas for growth in the marketplace.Networking Pipeline: What are some of those short-term areas?

Estrin: One is the whole area of [tying] management and security together. People have already put in the bandwidth. How do we really manage and secure that network? The model for security has changed. It's no longer just a firewall and keeping outsiders out, it's much broader now, in that you have to protect from authorized users, because of worms, because of contractors moving in and out.

The whole security landscape, and architecture, needs to be different. There are new threats that come from the inside, as well as from mobility. Mobility of contractors, employees -- it all impacts how you do your management, how you do your security.

A second area is cost. People are more price-performance sensitive than they've ever been. The question is, and has been for a couple years, what will happen to the Dells in the switch business? Will there always be a premium put on switches, are the wiring closet switches different than the edge switches? What is going to be the dynamic in the different parts of the switch market, between Cisco and HP and Dell? At a time when people are more cost-conscious, is there more of an opening for a Dell, or inside a Cisco, with Linksys, potentially competing with other [Cisco] products?

On the other hand, that price sensitivity is countered by this need for management and security. Management and security tends to mean you've got to have a certain set of features built in, so...Networking Pipeline: ...Do you pay for the cost in the box...

Estrin: ...Or in the managing of it. That will be interesting to watch.

There are a couple other areas that are driving different requirements for the network: Voice over IP, and storage. One of the interesting side effects of both of those is latency sensitivity. Up until now, most switches and routers haven't had to worry about latency. It didn't matter, it was about bandwidth.

But when you look at some of these new [network] uses, latency becomes important. One of the question marks is: Are the current incumbents going to use that [latency performance issues] to continue their incumbency, or is that an area that somebody new can target and compete?

The last short-term item is looking at the first inch, as opposed to the last mile, of the network. As in, where the server connects to the network. We have become so IP-centric in everything we do, that data center architechtures have evolved. There's a move to lower-cost Linux and multiple servers, and distributing applications on multiple servers, which means lots of communications between the servers.We don't have an efficient server I/O architecture today. There's things like InfiniBand, but people really want Ethernet. So it's really about looking at the way a server processes the packet, in and out of that server, with Ethernet and IP and doing that in a way that is efficient, high throughput and low latency.

We think that's an interesing area, and that's the domain of Precison I/O.

Networking Pipeline: What's interesting, looking farther out to the future?

Estrin: Home networking, and the consumer space. Its impact on networking isn't about fundamental architecture change. But home networking has been talking about the last mile for ages, and we still don't have it. Mobility, and that means beyond just connectivity, but services mobility, where you can really go to a hot spot and it works, and it's easy to use. Some of this is outside of the enterprise IT domain, but over the next three years it will impact IT folks, because whatever happens in the home ends up feeding [the enterprise], because people want ot use the same environment.

Looking five years out, the most exciting thing is the whole move toward embedded computing and sensor nets. Technologies like RFID are just the beginning. That type of networking will take a whole new architecture to support low-power devices, with lots of communication. I think you'll see that come out of academia in about five years or so.Networking Pipeline: We've been watching Washington a lot lately, what with all the talk about VoIP regulation. Should enterprise IT professionals pay attention to politics?

Estrin: You absolutely have to pay attention to it. At Cisco, we had a technology policy group in the office of the CTO. The reason it's important to pay attention is that what happens from a regulatory standpoint can dramatically impact cost.

One thing you learn about Washington -- the tide may seem to be going one way, but until it's done, it's not done. It's hard to know what the outcome may be.