The San Diego harbor is a warm and beautiful place. A large body of water instills a sense of calm in me. Yet as ships slither by, jets land and take off from the airport across the water. This juxtaposition of steady progress and nearly chaotic coming and going reflected many of the discussions during the first Net Database Summit held in San Diego December 3-6.
Web technology has made many developers rethink how they build database applications. But as is typical of the early stages of any sweeping change, controversies abound. As the conference chair, I was actually pleased to see our expert speakers publicly disagree with each other. Can current Web technologies really support high-volume transaction processing for more than 1,000 concurrent users? Some say, "Yes," and others say, "Not yet, but soon."
The Java steamroller has not yet flattened every obstacle in its path. Despite the excitement surrounding Java, many of the attendees I spoke with said their Java development efforts are still in the exploratory stages. And even those who have gone full steam ahead acknowledge that Java development tools are far from mature. They are better this year than last year but still leave much to be desired.
The importance of architecture was a persistent theme throughout the conference. One concern I heard voiced often is that modern multitier architectures, especially those based on Web servers and browsers, contain many more moving parts than simpler two-tier client/server architectures. That means there are more potential points of failure. This may be the seed that sprouts the next alternative approach. A major justification for multitier architectures is the promise of improved scalability. But as Ken Rudin noted in his presentation, "Building OLTP Applications for the Net," all parts of a system must be scalable, or the system as a whole will not be scalable. To me this means that the more parts we have, the more likely one will not be scalable enough, and the whole house of cards may come tumbling down.
Whatıs the inevitable result? New trends usually are driven by and seek to overcome the weaknesses of preceding approaches. If the fatal flaw of todayıs architectures is too many tiers, then simpler architectures with fewer tiers will probably enjoy a resurgence. I doubt we will rebound all the way back to old-fashioned client/server or mainframe host-based architectures; the pendulum will probably settle in somewhere in the middle. As David Linthicum noted in his presentation, "Application Architectures for the Net," the functions traditionally performed by separate transaction processing monitors are showing up as embedded functionality within other middle-tier application servers that have absorbed processing formerly done on clients. I think this trend will continue, and todayıs four- and five-tier Web architectures will be seen as too frail and complex to administer over the long haul. Expect to see more functional consolidation occur in the middle ground.
Architectures are important, but tools make an architecture come alive. In a Web-based architecture, the client browser is the front line for end users. Several people I spoke with are upset about the diverging functionality supported by todayıs browsers. The early hope that developers would finally have a widely adopted and standard platform for application building is waning. It may seem that we have only two vendors to consider supporting, but the rapid pace of upgrades has left a growing problem of version incompatibilities in its wake.
A good learning experience is even better with a good laugh or two. The funniest moment occurred as early as the second session of the conference. Neil Raden was giving his fascinating talk, "Pushing the Data Warehouse," when all of a sudden the telephone on the table next to the podium began ringing. We placed a phone there so speakers could dial out to the Internet as part of their demonstrations. I neglected to turn the ringer off, and my heart skipped a few beats. Neil turned his head, picked up the phone, furrowed his brow, then exclaimed, "I told you never to call me here!" and hung up. Thatıs thinking on your feet, Neil!
The Net Database Summit is part of the Database Summit series produced by Miller Freeman Inc.ıs Database Group, the publishers of DBMS and Database Programming & Design. The Data Mining and VLDB Summits are up next in Beverly Hills, California, March 1-7, 1998. For more information, please check www.dbsummit.com.