Status checking for Kaiser Permanente consists of more than pinging an object and waiting for a response. Tuttle needs to know that an object is not only alive, but alive and well. For failover protection, he runs two instances of the Oracle Parallel Server engine using a shared-disk configuration. He keeps most of the objects on the application servers and the servlets on the Web servers--a typical configuration.

The problem is knowing that all objects are working properly when there are multiple copies of the objects, multiple redundant servers, and a parallel database engine that allows a single connection. Testing these configurations from outside the firewall adds an additional complication, but one that Tuttle cannot afford to overlook.

CMGI's AltaVista Live, a personalized portal that builds Web pages dynamically, faced the same problem. "Deployment across numerous servers is a difficult issue, but subsequent checking that all components are installed properly and running correctly is the real challenge," says engineering director Allan McNaughton. "We have to monitor our setup all the time, because a portal cannot go down." To do this, he relies on test URLs that exercise all the components he wants to monitor. McNaughton uses SiteScope from Freshwater Software to do the testing and monitoring. The software reports any problems it encounters, including inordinate delays. "This tool helps, but there is a lot more to be done in this area," McNaughton says.

With large, multiplatform applications, interobject communication issues can create another management headache. Previously, components based on competing object technologies could not talk to one another. Fortunately, this problem has begun to abate. Corba and Enterprise JavaBeans use the same communication protocol, the Internet Inter-ORB Protocol. Hooks to Microsoft's COM/DCOM from Java and Corba also exist, so a Corba back end can pump data into a Microsoft Excel spreadsheet.

Rand Merchant Bank in Sandton, South Africa, uses interobject communication technology to cut down the amount of time developers spend translating component standards. "At first, we were writing our own wrappers, which was a workable solution, although a poor use of our resources," says Vincent Coetzee, Rand Merchant's chief systems architect. Rand Merchant employed the Object Bridge from Visual Edge, which automated the interobject communication.

An equally pressing problem is the communication among distributed applications and enterprise resource planning packages. According to the 1999 Technology Forecast from PricewaterhouseCoopers, only Madrid, another product from Visual Edge, performs such a conversion. Enterprise application integration software should allow communication among ERP packages and component-based distributed computing products.

Of the various object-management problems, communication across component architectures is the one most likely to be resolved soon. That's because nearly every vendor has a high stake in having its objects and components work with the enterprise infrastructure in place at prospective customer sites.

But the management of distributed components in the enterprise suffers from far more problems than solutions. The Web has pushed many companies into distributed computing before the paradigm could support the demands made of it, and many sites would confess to a disquieting feeling that they are vulnerable. Until better tools and services emerge, expect to read more about important sites going down. Just pray it's not yours that makes the news.

Andrew Binstock is the senior analyst at Pacific Data Works LLC, a firm that specializes in market analysis and technical white papers. He can be reached at alb@pacificdataworks.com.

return to page 1, 2