"If there was any single act that really helped Cisco, it was the merger of Wellfleet and Synoptics."

May 27, 2004

11 Min Read
Tony Li

53523.jpgTony Li has had an uncannily successful career path: Cisco Systems Inc. (Nasdaq: CSCO), Juniper Networks Inc. (Nasdaq: JNPR), and then Procket Networks Inc. OK, so, the jury’s still out on Procket (see Procket Processes a Dream Team and New Kid on the Cisco/Juniper Block). But the fact that Li helped build two of the leading companies in the router market is more likely a reflection of his skill and intensity in crafting router code, rather than luck.

Li hopped on board Cisco systems in January 1991, a short time after Cisco’s IPO. There he stayed for five years, becoming one of the company's most accomplished and senior coders. When Li joined Cisco, there were 275 employees and $69 million in revenue (in fiscal 1990). By the time Li left, there were 11,000 employees and $10 billion in revenue.

Next stop? Juniper, where Tony helped the founding team put together the first legitimate challenge to Cisco’s routing leadership since Wellfleet systems, which was ultimately gobbled up by Bay Networks and later Nortel Networks Corp. (NYSE/Toronto: NT).

Adding to his reputation as somebody who marches to the beat of his own drum, Li left Juniper just prior to its IPO, in 1999, when it seemed to be firing on all cylinders. He left to move on to another startup, Procket, where he logged another five-year stint, weathering some pretty serious melodrama before quitting earlier this year (see Procket, Procket Bulks Up, Procket Gets Unstealthy, Procket CEO Resigns, Procket Plows On, Procket Gets Cisco Exec, and Li Finally Quits Procket). Well, we’re still waiting for the Procket IPO. Maybe Li's departure has cleared the way.

We interviewed Li at the Peninsula Creamery, located in sunny Palo Alto, Calif. Unfortunately, both Light Reading editors were late. We were worried that Li might flash some of the mercurial temper he’s rumored to possess, but there was just this calm, relaxed man sitting greeting us with a boyish, slightly nervous grin.

As the inquisitors, we attempted to pry back the protective layers and draw out the inner flame of the “Li-ster,” as he’s known in certain circles (well, okay, we just made that up). No such luck. Li seemed protective and ultra-wary of the Light Reading pen. And he really just wanted to talk about the Internet. Even then, his words were measured. Leisure and idle conversation do not appear to be high on Li’s list of priorities.

That’s our way of saying, this is a short interview. [Ed. note: But we're making up for it with this long-ass introduction.]

In the end, Li’s own description of himself as “workaholic” seems most apt. He’s a guy that would rather be designing, implementing, and writing routing code. Just take a gander at his voluminous posts on the Light Reading Message Boards, and you get a pretty good idea that he lives and breathes what he does. How’d he get there? It’s an interesting story.

Li went to Harvey Mudd College, in Claremont, Calif. He proceeded thence to USC to earn his PhD. At USC he started fooling around with these new-fangled networking boxes called routers. At the time, he was working in a computer lab, and he got involved configuring routers made by a company called Cisco Systems Inc. The rest, as they say, is history.

Li was mum on the behind-the-scenes drama at Procket, though he did drop a few hints. He’s currently in “semi-retirement,” although its clear that doesn’t mean time on the Links.

Read on, to find out what Li thinks about Cisco, the Internet, telecom, and startup life:

  • Cisco Kid

  • 50 Ways to Leave Your Paycheck

  • Virtual Liechtenstein

— R. Scott Raynovich, US Editor, and Craig Matsumoto, Senior Editor, Light Reading

Light Reading: So tell us how you got started at Cisco.

Li: I started Jan 14, 1991, after the IPO. This was the week that the Gulf War started. Nothing like having a bit of excitement at work. Cisco was my first job after grad school; there were 275 employees.

My PhD was in computer science. More specifically, compiler theory. Even more specifically, in compiler correctness. I learned about Cisco routers at USC. I was working at the University Computing Services group as a Unix sys admin and net admin. I was "selected" to help with the initial launch of Los Nettos, and I was USC's technical contact for that. The network was built from Cisco AGS systems with 4 T1s. At the time, it was incredibly fast. I was one of the router jockeys involved in that network. The leadtechie was Walt Prue, who is still there, doing the same thing.



At Cisco, I first worked in the IP routing group, and I spent most of my time on maintenance. I’d get a bug report, I’d fix it, submit a patch. I did that for a few years. Then I worked on BGP. I did the BGPv3 implementation and then I helped spec out BGPv4.

Light Reading: You hear a lot of doomsday stories about BGP being at the end of its life cycle. What do you think?

Li: BGP is capable of carrying a certain amount of load given a certain amount of processing power. It can scale to a certain level with today’s hardware. If the size of the routing table grows faster than the hardware, BGP runs into trouble.

Light Reading: Is that happening?

Li: Not yet. But it’s easy for it to happen. IPv6, if not deployed carefully, can grow the forwarding table by a factor of five. This happens if we don't have careful prefix aggregation and we get the same number of v6 prefix that we have today.

Light Reading: So how big an issue is this?

Li: The size of the tables depends on how the Internet community injects routes into the Internet. It’s an ongoing matter of vigilance on the part of service providers.

Light Reading: You were at Cisco for much of its rise to dominance. Any reflection on how Cisco won?

Li: If there was any single act that really helped Cisco, it was the merger of Wellfleet and Synoptics [to form Bay Networks, which was later acquired by Nortel].

Light Reading: So you think the merger didn’t work?

Li: Not at all. It slowed them down. There’s no doubt in anyone’s mind that Wellfleet had better hardware. But the overall system is controlled by software. That’s why it’s called a system, and it’s about who has the best system.

If you look at the basic history of Cisco, it got in on the ground floor when there were not other router competitors. Proteon, Wellfleet – they were peers and competitors. There was an open field at the time and then the merger happened.

Light Reading: So what is it with all these stories about how you left Cisco? The story of nailing a letter to a door...Li: It’s a long story.

Light Reading: Sounds like a short story, actually. But is it true, about the nail, or the knife, or whatever it was?

Li: Let’s just say I’m not proud of it.

53523_1.jpgLight Reading: So nothing like that happened at Juniper.

Li: I left Juniper very quietly.

Light Reading: That’s not what we heard. We heard you left on the eve of the IPO, or something like that.

Li: Yeah, but that’s a completely irrelevant coincidence.

Light Reading: What was the highlight of working at Juniper?

Li: I'd say the real highlight for me was being at a startup, free from all of the corporate overhead that Cisco was, and is, embroiled in. I think that they're doing a fine job competing with Cisco in the carrier space. How they will fare in the enterprise space remains to be seen.

Light Reading: So then you ended up at Procket. Tell us how that started.

Li: It was originally started by Sharad Mehrotra. He was the initial guy, and he was trying to get funding. Then he brought in Bill Lynch. The original idea was to build a network processor. Both of their backgrounds were in chip development. I was introduced to them because of people I knew, including Stu Phillips at U.S. Venture Partners. I explained to them that they were very close to building a router. One thing led to another, and we realized we were capable of building an interesting system.

Light Reading: You were at Procket for quite a while. Why did you leave?

Li: Some of the things going on were incompatible with the way I thought that things should be done.

Light Reading: Engineering-wise or operationally?

Li: More of a company culture, business practices.

Light Reading: Did this begin when Roland Acra came on board?

Li: That had some effect. Light Reading: Do you think Procket is still the right product, or was it built for a different time?

Li: It’s still an excellent product for the core market. It has industry-leading technical capabilities, and I’m very proud of the system.

Light Reading: But folks aren’t buying many core routers these days…

Li: That’s part of the issue. The market is no longer what was expected. Certainly the capex budgets are not what they used to be.

Light Reading: We checked out your Website. Nice photo! How’d you get “tony.li” its own domain?

Li: The Republic of Liechtenstein. All I had to do was figure out that the two letters of my last name, L-I, is the country domain for Lichentenstin. Anybody can do it.

Light Reading: Well, not everybody. You need a two-letter last name. "Craigmatsumo.to" might not work. So you never actually had to go to Liechtenstein?

Li: No. They gave their domain registration to the Swiss, and it’s all on the Web, automated. Hand them the credit card number, and they’re very happy to do it for you.

Light Reading: How’s the job search going?

Li: Pretty good. I was looking around for ideas on your Website.

Light Reading: Finding anything? Li Really, I think I’m looking to stay low. You know, something to keep me out of the house. I’m pretty much retired. But for me, that’s different from when other people say they’re retired. I’m a workaholic of the worst order.

Light Reading: What do you do in between jobs? Do you hang out at Starbucks and write router code?

Li: Consulting. Helping in some other places. Basically hanging out and seeing what else is going to happen.

Light Reading: So would you join another startup?

Li: If it were the right circumstances. If it were the right people and the right idea.

Light Reading: You could always work for us, write about routing.

Li: [Pauses, as he chews his quesadilla.] Actually I hate writing. I just hate it.

Light Reading: But you write on our message boards a lot!

Li: That’s different.

Light Reading: You’re right – it means we don’t have to pay you for it. So tell us about being an engineer in Silicon Valley, post-bubble. Do you think good engineers are really struggling to find good work now? Are good people moving away?

Li: I don’t have trouble getting jobs in the downturn, upturn whatever. Certainly it will be easy to attract talent. But I don’t see a major trend or exodus.

Light Reading: What about the outsourcing trend?

Li: Well, I’ve had my head down in the last few years. I haven’t seen a lot.

Light Reading: We’ve heard that’s the new startup model – have half your engineers in India, and have your marketing here.

Li: I can’t say I have any expertise in this area. But I don’t think that’s the obvious thing to do. There are definitely pros and con. The pros are you save on labor cost. But you put yourself in the situation where you are managing a geographically diverse organization and you are opening yourself to quality control issues.

Light Reading: What did Procket do with outsourcing ?

Li: Procket has a very small, opportunistic outsourcing program. It really is more housing existing Procket engineers there than anything else. There were employees that wanted to go back to India, and they wanted to stay with the company.

Light Reading: Do you think that optical routing is possible in the next few years?

Li: Not without a significant breakthrough in buffering. These aren’t things the optical domain is particularly suited for. At the end of the day, you need a router that is competitive on price with an electrical system. If you are using optics or not, it doesn’t matter if it’s not competitive on price.

Light Reading: The dynamic of service providers has changed quite a bit. It used to be ISPs running the Internet, now it looks like it will be RBOCs. What do you think of the future of service providers?

Li: The real question is whether an independent ISP can compete with RBOCs and the monopoly of the last mile. If there are technical solutions that allow us a new last mile, there are possibilities. Cable, for example, has had significant penetration. There’s other players. I’m hopeful that we’ll see a wireless last mile.

Light Reading: Describe the technology.

Li: The requirements are pretty simple. It needs a range of 20,000 feet and it needs to be dependable in a reasonable manner with reasonable throughput. I don’t know that wireless can do that today, but that’s not my area. But I’m sure somebody’s working on it.

Subscribe and receive the latest news from the industry.
Join 62,000+ members. Yes it's completely free.

You May Also Like