Start-up Ipsilon Networks is at the heart of a fierce religious debate over how networks are constructed.
The networking hardware and software maker has introduced IP switching, a concept that shifts the performance responsibility on a network away from the router, the device that has made Cisco Systems the giant of the networking industry.
Using the dominant communication protocol of the Internet, IP, and layering it on high-speed ATM (asynchronous transfer mode) technology, Ipsilon has forged a new method to speed data to its destination without the need for a router, which is viewed by some as a bottleneck due to its use of complex tables to route data to its destination.
Others have followed in Ipsilon's wake. Cisco has introduced a concept called Tag Switching that incorporates routers in an IP network. In addition, 3Com unveiled a strategy called Fast IP that incorporates some of Ipsilon's technology. Neither of those IP switching offerings is shipping. Others have delivered a pure hardware solution that offloads IP-related data processing from the router in order to reduce congestion.
CNET's NEWS.COM recently sat down with Ipsilon founder and current chief technology officer Tom Lyon to discuss the state of his company and the role IP switching technology can play in networking. Lyon, a veteran of 12 years at Sun Microsystems, started the company in October of 1994.
NEWS.COM: What was the impetus for the IP switching concept?
Lyon: At Sun, I was very active in the ATM community and, of course, active with IP, since that is the heart of Sun's networking. I saw two major trends going on. First was the incredible growth of IP. Then in the ATM community, ATM hardware was turning out to be really incredibly fast and pretty amazingly cheap, too. But the two had nothing to do with each other. The ATM Forum was marching down the path of the ISDN, OSI-style protocols and they had made a decision fairly early on--that I was part of but did not agree with--of inventing this new world instead of trying to copy IP. Because of the lack of good performance in the router world and with incredible dominance of IP, I saw a good opportunity to bring ATM and IP together.
Where will Ipsilon focus on expanding its product line?
There's definitely OEM (original equipment manufacturer) possibilities. Right now, our IP Switch Processor is based on a Pentium PCI device, so it's very easy for us to incorporate PCI hardware options, so you'll see a lot more of that. Also, we'll probably start to OEM a few boxes for various kinds of edge devices.
Ipsilon has criticized the role of routers in a network, saying those types of boxes cannot scale. What is your reaction to the upcoming introduction of the new 60-gbps router, also known as Big Fast Routers (BFR), by Cisco?
I never had any doubt that Cisco could build a very high-performance router, but if you look at price/performance, there's no incentive for Cisco to improve the price/performance of routing. They just leave money on the table if they make stuff cheaper. The BFR will have incredible performance grading, but no one will be able to afford it. So the Cisco 7500 routers are still going to be the volume product. We're going right at the 7500 saying: "There's a lot faster and cheaper way to do this."
What do you think of some of the alternatives being floated to the market, like Cisco's Tag Switching?
[It's] technically very similar to IP switching, but again, they're oriented toward the very highest end of the market. They don't want to disrupt anything going on in the volume market. The other key thing about Tag Switching is that I don't believe Cisco is serious about making it work with ATM. What they want to sell is BFRs. We want to work very well with ATM so we can partner with all the ATM vendors in the world.
Do you think your reliance on ATM will limit you in the networking market?
You'll see our IP switching over Ethernet sometime. Right now, we're awfully busy serving the ATM market. It'll be driven by various partnering opportunities.
How about 3Com's Fast IP concept?
What 3Com announced with Cascade Communications, Novell, and IBM is that they would use our IFMP protocol for interoperability among each of their proprietary schemes. Fast IP is interesting but it seems to take a lot of the concepts from the ATM world, about end-to-end quality of service, and that stuff and tries to move it into the Ethernet world. But those concepts were never really proven in the ATM world so I'm a little skeptical of their success in the Ethernet world.
What is your view of the standards debate? Opponents are quick to label Ipsilon's scheme as a proprietary technology.
It's proprietary, but it's absolutely open...There's standard things that are among the most closed systems on the planet. Standards are something that are very poorly defined--the ATM Forum, for instance. If you look at the ATM Forum charter, it is not a standards body. Specifically, when you join, you agree you are not there to create standards. So is the ATM Forum's MPOA a standard? It's only a standard because there's bunch of vendors doing it. It's the same with IP switching; there's a bunch of vendors doing it now.
How do you differentiate yourself from competitors?
The major differentiator we have is that we have product and it works. That will continue to be the major differentiator. We're not our there making wild claims that are unsubstantiated; we have product. You can buy it, you can try it...Other people are just way behind the curve.
Where does Ipsilon's customer base stand as of now?
Probably around 100 on a worldwide basis. There's probably 20 percent now that are in production networks. People tend to go through at least a six-month evaluation test period before they really put in a backbone, though we're seeing that cycle speed up quite a bit now.
What is Ipsilon's role in a distributed network?
IP can go into different environments. ATM can also go into different environments based on what physical speed you are using. We're already starting to deal with partners in the DSL space and looking at getting IP switching all the way to the home. Again, we don't claim that to be a standard, but if you look at what's standard, there's nothing there. It's a total mass of confusion, so we can offer those guys something that works today.
Ipsilon gets a lot of credit for creating the IP switching concept, but how long a start-up can survive now that more established players have entered the market?
We're going to continue with partnerships to help us from a market position and also from a cash position. Cabletron Systems invested $20 million. It serves us well in terms of continued growth. I don't see us slowing down at all.
Is the goal to eventually go public?