Umang Gupta has a unique perspective on the network industry, having been at the forefront of the client/server application market as the founder of Gupta Corp. (now Centura Software) and now as CEO of Keynote Systems, a company best known for its Web site performance and benchmark services. He spoke recently with Network World News Editor Bob Brown about the future of Web-based applications and the Internet itself.
What's your take on Web services?
Long term, they are a logical way for the world to go for a large class of applications, especially consumer-facing applications, but even interorganizational ones. How close they are is a different matter. There is a class of Web services we know are coming from Microsoft -- Passport is a classic example that's already here. How they emerge from everyone else has only partly to do with the technology. A lot of it has to do with the right business climate. The analogy I would use was the original introduction of Windows in the early '90s. People often wondered what applications would emerge under Windows to replace the old character-mode applications, and while the initial applications were the classic spreadsheets and word processors, today of course there are thousands of GUI-based applications. The same level of unleashed creativity is going to determine how big Web services finally get. Once the general technology, standards and [software development kits] are available for people to build new classes of Web services, they will be everywhere.
Fusion-exclusive questions regarding VPNs, convergence and more.
Is this going to come down to another Windows vs. Java battle?
Ultimately it will end up being a form of a standards battle, but the bigger question is not Java vs. the Windows platform. The bigger question will wind up being desktops vs. devices.
How so?
The desktop battle is over and the winner is Windows regardless of what happens with the antitrust cases. Whatever Microsoft offers, whether it's [Internet Explorer] with or without Java, that's what's going to dominate on the desktop. But to the extent that desktops themselves over time will end up being replaced or augmented by other devices, the battle is far from over as to what platform software will sit on these devices. If it's a phone it could very well be based on Java, Microsoft or even Nokia stuff. If it's a PDA it could be a Windows platform. But I think Java could have a shot also. Web services ultimately will succeed because there's a class of things people want to do with devices that are much better done with Web services than with software sitting on their devices. Location services for wireless are a very logical Web service. So I don't think in that particular case it's necessarily going to favor Java or Windows. It'll depend on whose got the best devices and services.
How will devices actually handle these potentially complex Web services?
We went from a client/server world which was largely a thick client and a thin server world, to the Internet world, which allows you to build around a thick server in thin client mode, even with diskless devices. With the next generation of devices and Web services, we'll be talking about thick clients and servers. On the server side, there will be a lot more intelligence needed to drive these Web services, but on the client side you're going to need some pretty smart software - browser software, device [user interface] software or what have you - that is capable of integrating data from multiple Web services and still presenting it logically to the client. These browsers themselves are going to have to get much smarter…we're already seeing that with [Internet Explorer].
How should network executives prepare for Web services?
Everybody's going to have to be making changes constantly. It's not going to be a one time revolutionary thing. The big changes on the client side will end up coming from a vendor like Microsoft with a new version of its browser or device vendors who will integrate thicker browsers into their [user interface] software. On the server side, people will have to choose sides I suspect. If you're buying NT servers and standardizing on them, my guess is you'll wind up with a lot of Microsoft standard services. If on the other hand you're buying Solaris or IBM or any of the other platforms that are potentially more scalable, you may get a different class of Web services. All the discussion about all of these companies working together is encouraging, but the devil is in the detail. We've all heard the talk of cooperation on standards for Unix for many years and we know how many variations of Unix exist.
What is affecting Internet or Web performance these days?
At a high level, it usually comes down to one of two things. Either the application is not well constructed or it's not well connected. What we've found over the last few years is that there was a large set of connection issues, so that you didn't even know if you had application construction problems. Four or five years ago people were largely concerned with issues such as: Do I have the right bandwidth supplier or Web hosting supplier. Today through, this is less an issue because customers are smarter about choosing the right vendors and because of the consolidation in the industry that has resulted in fewer but bigger and more reliable players offering these services. Now there are more issues emerging at the application level [and Keynote is evolving its product line to address such things as application-level performance and diagnostics].
Are content delivery networks having much of a positive impact on Internet performance?
In general they work well for static data. I'm not convinced, however, that's good enough. Most sites are 90% dynamic and 10% static, and over time they'll be 99% dynamic and 1% static.
How indestructible is the Internet?
I'm a firm believer and all our data would show that it's pretty indestructible. Ultimately, there are always some outages here and there. Even if something hit MAE East or MAE West [large Internet switching centers] I'm not sure it would make a significant difference for too long because first of all the Internet self routes. The Internet never falls apart, what happens is it simply starts to degrade in performance. All the data we've seen about performance whenever outages have occurred, on the east or west coast or anywhere else, is that they're usually localized problems to some backbone or geographic location and they are limited in time.
What's your take on VPNs?
For many companies except for the extremely large companies that can afford dedicated WANs, the most cost effective way will be VPNs. After all, you are piggybacking off a public switched network and public standards-based Internet, so ultimately the cost differences are so huge that VPNs are the better way to go for most people. On the other hand, you do give up a certain level of reliability and possibly performance -- and to some extent security. But there are a lot of companies in that area working to make improvements.
What impact will VPNs have on the Internet?
I'm not sure. So much investment is being made on the Internet in terms of the backbone, fiber capacity for the underlying transport layer and software for adding reliability at the application layer itself that I'm not sure that by themselves VPNs will make a significant difference to Internet performance.How about voice and data convergence?
People are starting to do it. We, for example, use Cisco's IP phones. Still, most of America is keeping voice and data separate. But having said that, would converged traffic affect the Internet as a whole in terms of capacity? You bet. Clearly when we move to a world where people have a 1,000 times more data than voice that's going to impact capacity. My sense, however, is that the problem will not be in raw bandwidth capability. There's enough bandwidth to meet the needs of most companies for the next decade or two. Witness the overcapacity and pain and suffering many bandwidth companies are having today. The problem, if there are capacity bottlenecks, won't be at the pure bandwidth infrastructure layer, but more at the router or switching layer involving the lines feeding into the core. That might need to be upgraded over time.
Recent reports have suggested that the Web is shrinking, at least in terms of registered domain names. Have you seen any sign of this?
I don't have any independent data that would support that. My guess is we're simply going through the typical withdrawal pains of a boom. But ultimately every human on earth could arguably have a domain name. I don't mean to be a Pollyanna on this, but it's only a matter of time before most people end up needing communications beyond e-mail on the Web and a Web site or domain name could be an important part of the addressing system for individuals.
RELATED LINKS
