|Home - CEO Spotlight - Jun 05 Issue
Entrepreneur Spotlight: Mike Stonebraker, StreamBase
continued... page 2
Angel Mehta: At the rate that countries like India and China seem to be gaining and enhancing their R&D infrastructure, is it possible for this side of the world to stay in a leadership position over the next 30 years given what the trend has been and seems to be?
Mike Stonebraker: Let me answer that question obliquely. Let’s look at a little statistic. Do you know about the Turing Award?
Angel Mehta: Yes, it’s like the Nobel Prize for computer scientists… right?
Mike Stonebraker: Sort of, yes. I guess it’s been given since 1971, so there have been 33 of them or something. Another award is the John van Neumann IEEE award given annually to those who the IEEE thinks is the world’s best computer scientist in the last 13 years; there are 44 of these people give or take a few….
Angel Mehta: Aren’t you slated to receive that one this year? [Laughing]
Mike Stonebraker: As it turns out, yes… I’m one of the 44. [Laughing]
Angel Mehta: Congratulations.
Mike Stonebraker: Thank you. Anyway, the point I was making is there are no representatives from Japan, China, or India who have received those awards over the last 30 years because at the top of the food chain is ‘idea’ generation’. The U.S. is just unbelievably strong at innovation – it’s the thing that differentiates us from the Japanese. I think the Japanese are very very good at taking existing ideas and making them incrementally better but they’re not good at creating, thinking out-of-the-box. As long as IT innovation comes from out-of-the-box thinking, the U.S. will be in very good shape. Eventually, India and China will get there – but they are a long way away.
I will also tell you that if we’re going to stay ahead of the Japanese, Chinese and everybody else, there has to be a way to infuse new ideas into United States IT companies. If they’re not going to do it themselves, the only way to make that happen is to have start-ups run with them, threaten the established vendors, and thereby force them to pay attention. I think the only reason the U.S. Information Technology sector is vibrantly healthy and is clearly beating everybody else in the world right now is because we have this brain trust in research institutions and lots of venture capital that will foster technology transfer, as well as numerous entrepreneurs who are willing to take those ideas and run with them. They end up threatening big companies and that way you move technology forward.
Angel Mehta: Can you give me one key lesson from each company that may have changed your approach to building the next one?
Mike Stonebraker: That’s easy. When we started INGRES, I didn’t know anything at that time and, therefore, we hired the wrong president. We just plain made a mistake. The problem is that it takes a while to figure out you made a mistake and then it takes a while to convince the investors that it’s a mistake and then say “Oh my goodness if we fire this guy and hire somebody else, it will take six to nine months and we’ll lose all this momentum”. I guess the lesson I learned from INGRES is if you’ve got the wrong guy at the top; cut your losses immediately and move on.
The lesson I learned from Illustra was very interesting. Illustra had a fabulous product but had a ‘chicken-and-egg’ problem. It required plug-ins that were written by other people in order to get market penetration. The other people looked at us and said, “You’re an alternate distribution channel for our software. What’s your total sales turnover and why should I be interested in you?” We had trouble building sales because we didn’t have plug-ins and we had trouble recruiting the plug-in vendors because we didn’t have sales, so we were struggling. Then along came the Internet and Illustra became the database for cyberspace and we rode the coattails of the Internet just fabulously; this was all because of superb marketing. So superb marketing is absolutely crucial to success but chances are you’re going to have to grab on to some brass ring as it goes by.
The third company, Cohera, was much less successful. Cohera was selling a federated database system. We would integrate multiple existing physical databases and make them look like they were just one. Again, it was a very elegant product but there is not a big market for federated databases. We were having a hard time building sales. In fact, the company was ultimately sold to PeopleSoft; the investors got some of their money back but it certainly wasn’t a financial winner. Early on, we realized there wasn’t much of a market for our product and we started morphing it to do things that were more useful. I think we should have shut the company down, returned the investors’ money and moved on. I learned that if your primary market isn’t there, trying to go into search mode to morph into something else is not a tactic that usually works. I think lots of companies that I’ve run across would probably have been better doing that and that would have saved a long slow agonizing trip into the sunset.
Angel Mehta: Cohera was ramping in the bubble years, wasn’t it? 1999 time frame?
Mike Stonebraker: Yes, absolutely. The pressure from everybody was “grow as fast as you can, buy market share and don’t worry about how much money you spend”. Of course, we all hit the wall in 2001 and 2002. I learned that every start-up should kiss every nickel and should be thinking… it’s a long desert out there and the next oasis is God knows how far away… you better conserve your water. That’s something we fervently do at StreamBase. I think being unbelievably frugal is the only way to run a start-up. I think companies that successfully survive the Internet bust in 2001-2002 were the ones who did exactly that.
Angel Mehta: One of the things that everyone seems to agree about, or certainly the investors I spend time with, is that software as a whole is consolidating and the real growth is over. Do you see any monumental shifts on the horizon for enterprise software that might cause a new wave of opportunities the way that the shift to client server or web-based architecture did?
Mike Stonebraker: Au contraire. I think enterprise software is alive and well. There was a huge boom in 1997-2000; the Internet drove every CIO to open up the purse strings and spend money on everything imaginable. There were a ton of companies started which no reasonable person, in retrospect, would ever have invested in. I guess it was pets.com that was going to sell pet food over the Internet… pet food is about as heavy as cement… and no one would sell cement over the Internet and ship it FedEx because it just doesn’t make any sense. Investors were just caught up in this feeding frenzy and most of them have gotten a big dose of humility since then. I think optical networking was another sort of boom-that-went-bust kind of technology. If you say what’s going to be the next big thing, well there’s no question that the declining costs of sensor tagging through a variety of technologies is going to cause everything of significance to be sensor tagged and reported in real time; this is going to lead to real time monitoring applications in all kinds of fields.
Angel Mehta: Hence the genesis of StreamBase?
Mike Stonebraker: Yes, in 2001, Stan Zdonik, the Computer Sciences Professor at Brown, and I realized that what I just said was going to happen and that was going to cause a fire hose of real time data to be processed by some downstream systems software. We realized that current system software was not up to doing this sort of application, so we set about writing a new stream-processing engine that’s good at doing exactly what I said. Over the next decade this revolution or sea change in sensor tagging will come into the marketplace. Let me give you some examples…
Let’s take California. The Bay area has an Easy PASS system. When you’re going over the Bay Bridge and you use the Easy PASS system, it reports your location, in real time, to California’s Highway patrol computers. You can also log into CHPs website and find out what is causing the delay on the bridge.
An evolution of that kind of assist is to do variable tolling. This concept involves charging you more at peak periods. So to cross the Bay Bridge at 8:00 a.m. will cost you $50 due to the horrible congestion in contrast to $0.10 to cross at 11:00 a.m. when there’s plenty of capacity. In my opinion, the only solution to the congestion on the Bay Area Freeways is variable tolling, extending that to Highway 101 between San Jose and Palo Alto and so forth or to Highway 680 between Freemont and Dublin.
That takes a massive system software real time infrastructure like StreamBase to accomplish. It may take a decade before we get to variable tolling on the freeway but it’s certainly going to come.
Angel Mehta: It could apply to family safety issues, as well, couldn’t it? Tracking your toddler for safety purposes….
Mike Stonebraker: Yes! What if your wristband at Disney World turns into a sensor? You’re kids are GEO-positioned in the amusement park at all times so there are no more lost kids.
Another example: MIT is deadly serious about putting a sensor tag on every parking space they own and turning my MIT ID into a sensor tag. They would then detect illegal parking without having to send a guy around checking people’s windshields. You can also keep track of all the empty spaces on the MIT campus directing people to empty spaces.
The Financial service community is completely driven by electronic trading. Everyone on Wall Street is doing it. I would describe electronic trading as a whole bunch of people looking over a construction site (‘the market’). There’s a bunch of bulldozers down on the construction site piling up the dirt and every once in a while one of them plows up a nickel. If you see the nickel, run and pick it up faster then the next guy you can keep the nickel. Electronic trading is all about latency and involves a high-volume war of who can figure out real time analytics faster then the next guy. We’re being very successful on Wall Street because of the sea change caused by electronic trading.
In network management, quality of service, denial of service is causing the network management guys to have to do high-volume analytic on their feeds. Again, the feed change is causing a shift.
Mike Stonebraker has been a pioneer of database research and technology for more than a quarter of a century. He was the main architect of the INGRES relational DBMS, the object-relational DBMS, POSTGRES, and the federated data system, Mariposa. He is the founder of three successful Silicon Valley startups, the objective of which was to commercialize these prototypes. Mike can be reached for feedback at: email@example.com
Angel Mehta is Managing Director at Sterling-Hoffman, a retained executive search firm focused on VP Sales, VP Marketing, and CEO searches for enterprise software companies. He can be reached for feedback at: firstname.lastname@example.org