Routing's rock star

Where Tony Li goes, success often follows.

Li, an Internet Protocol routing engineer, is a Silicon Valley wunderkind. He helped Cisco Systems develop its first true core IP router and then played a key role building Juniper Networks from a scrawny little start-up to the industry's No. 2 IP routing company.

Li left Juniper in 1999 just before the company's lucrative initial public offering. He then helped found another routing start-up, Procket Networks, where he served as the chief scientist. After a tumultuous five-year stint at Procket, Li resigned from the company earlier this year.

Cisco and Juniper flourished during--and after Li's--tenure. Procket, which raised over $300 million in funding, will be bought by Cisco for $89 million, the companies said Thursday.

Many people regard Li as one of the brightest minds in the IP networking industry. A self-proclaimed perfectionist, he acknowledges that some colleagues find him "difficult," which could explain why he has left all three of these companies amid some controversy.

CNET News.com caught up with Li shortly before Cisco announced it would buy Procket. During the hour-long phone interview, Li, speaking from his home in California, dished the dirt on his previous employers, explained his rock-star persona and gave some insight into new trends he sees in the IP routing market and the Internet in general.

Juniper is the only core routing vendor to successfully take on Cisco. How did they do it?
Cisco had annoyed its customers because they had not presented them with a reliable and useful core router. Cisco's corporate bureaucracy was structured so that people were awarded for playing very safely. At the same time, the Internet was growing by leaps and bounds. We were basically a roadblock in the growth, because we could not get a reasonable system to market. It is not that Juniper succeeded so much as Cisco failed.

You left Cisco and went to Juniper, which was then a start-up. Then you left Juniper in 1999, just before the IPO. Why?
It was time for me to move on. I had run into some issues there, and it was simply time for me to find other opportunities.

Didn't you leave a lot of money on the table in terms of stock options?
I left some things on the table, but it was hardly enough that I would care. I am one of these strange people who doesn't think this is all about the money.

Procket has raised something like $300 million, and now there's talk that the company is trying to sell itself. Could it really be out of money?
Unless they have raised more funding that is not publicly known, then I'd say they need to do something quickly. They should be close to running out of money by now.

People have been saying that Cisco is interested in buying Procket. Does this make sense to you?
I'm not sure what Cisco's rationale is. If it's a matter of improving hardware density and getting some software talent, then that makes sense. But I don't see how this wouldn't really (anger) the team that just built Cisco's new CRS-1 router.

People say that Procket could be going for between $80 million and $100 million. What's your take on that?
Well, I'm a bit biased, but I think that's a real steal for the technology.

Procket was one of the most highly valued start-ups in Silicon Valley at one point. What happened?

I am a perfectionist, and I certainly have a temper.

There was a lot of mismanagement, especially in the early days. Things were delayed, and the sales didn't happen. The product ended up being much more expensive to develop than we had planned, simply because we were not able to execute as we had hoped.

Were there partnership opportunities that came about and were they missed?
A number. But I can't go into them specifically.

There's been a lot written about the fact that you and the previous CEO, Randall Kruep, didn't get along. Is there anything to this?
I tried very hard to get along with Randall. That was basically rejected out of hand. I had nothing to do with him leaving. I did not want him to leave. It was actually sad to see that happen.

Why did you leave Procket? I thought that you were instrumental in getting Roland Acra, the new CEO, to come on board.
Once he got into the company, he wasn't interested in having my help. That was unfortunate.

Your reputation is that you are a difficult person to work with. Some people have called you a "hothead."
Well, I am a perfectionist, and I certainly have a temper. Sometimes that temper is not well served, and I guess I have the ability to irritate people. If people are not willing to have a rational discussion with me, then things get very, very difficult. I am always pushing to help the company succeed, but some people feel their egos need to be served.

You are considered one of the leading engineers in the routing community. Venture capitalists will pour millions of dollars into a start-up just because you're involved. Why do you think you are viewed as such a rock star?
First of all, I am not a rock star. A lot of this publicity has been internally generated. I have had no control over it, and I certainly do not deserve it. I am an engineer. My first goal is not to sell product, it's to make the customer happy. And I do strange things, like tell the truth, which customers appreciate but sometimes the company does not.

Now that you're done with Procket, what are you working on?
I am doing stuff on the side, but it is not of relevance to this conversation.

Do you think you will join another start-up?
Quite possibly; I would much rather be in a start-up than a large company. My style is much more about getting things done, and I prefer the freedom rather than the many layers of process that are usually necessary in a big company. I don't have anything particular in mind right now, but I am keeping my eyes open.

What are customers looking for now in a core IP router?

The question is really who is going to have sufficient technology as the Internet continues to grow.

Customers need a system that they can install and have operational in their networks for five to 10 years. That means you need an architecture that is going to be able to sustain Internet growth rates for 10 years without changing the basic infrastructure.

What do you think about Cisco's new core router, the CRS-1, which was just introduced last month? The company boasts it can scale to 92 terabits per second.
They missed the mark. Although it's significantly bigger than I had originally thought, it should still be an order of magnitude larger. Customers that have vision and understand the growth of the Internet will probably be disappointed with this product.

Right now the market is served by two vendors: Cisco and Juniper. Is there room for a third player?
Certainly. If you take a look at the traditional telecom market, we have always seen a number of players. The question is really who is going to have sufficient technology as the Internet continues to grow.

What's the next big trend in IP routing?
As the Internet grows, we'll see some very interesting things in the optical plane. I'm not talking about optical routing so much as optical traffic engineering. We've discovered that some traffic engineering methods for IP translate directly into the optical layer. This would allow carriers to deploy and provision optical circuitry at a very rapid rate to reflect changes in traffic demand.

Would this save carriers money?
It should make it cheaper to operate their networks, and it could provide a business advantage over competitors. Right now, if a customer wants an OC-48 circuit, he has to wait almost six months for it to be provisioned. Time is money, and if the carrier could get that OC-48 (a 2.5-gigabit-per-second optical connection) up in an hour, that might be a business advantage.