InformationWeek.com

Welcome Guest. | Log In| Register | Membership Benefits

November 15, 1999

Windows DNA 2000:
Distributed Development
Enhanced services in Windows 2000 make it easier to extend your COM code into the world of distributed applications, letting your business finally enjoy the benefits of applications that integrate many hosts

By Logan Harbaugh

Related links from our sister publications:

Computer Reseller News DNA 2000 Late To Web Race

Computer Reseller News Microsoft prepares to throw Tomahawk

Computer Reseller News New Era Of Application Management Emerges
or much of the last 20 years, application development seems to have evolved in circles. First, the PC revolution wrested control of many applications from centralized processing on mainframes. Then, more than five years ago, many companies began to embrace client-server applications, where part of the application, principally just the user interface, was executed locally on client systems, while a back-end server did much of the heavy lifting. At its core, this move to the client-server model was a return to the centralized processing of the Big Iron days.

In the past few years, however, IT shops have begun deploying applications that distribute their work over more than two systems. This technology is the foundation of middleware services and other object models, such as BEA Systems' Tuxedo and Sun's Enterprise JavaBeans. But it is the improved support in Windows 2000 for these multiserver applications--namely, the Windows Distributed Internet Architecture (DNA 2000) and the COM+ Distributed Component Object Model--that may well spur their adoption among more companies.

In its simplest form, this more distributed model lets IT shops write user-interface code that executes on the client like a conventional client-server application, but divides remaining code into a middleware layer, which includes the business logic of the application and a database layer that handles the back-end data-processing chores. In some applications, these three processing layers may be subdivided even further, with multiple servers handling different components of the business logic or multiple back-end servers managing different types of data. Indeed, any application with more than the two tiers typically found in a client-server model is known as an n-tier application.

N-tier models offer a host of advantages, particularly for organizations that want to expose core business applications to a variety of clients, including Windows desktops, Web browsers, and CE clients. By encapsulating business logic in a middleware layer that is separate from the user interface, IT shops need only write and maintain the logic of an application once. Although the clients may still require different applications, IT doesn't have to re-create, revalidate, or maintain diverse instantiations of the application's business logic. N-tier applications also scale better, because IT can improve performance simply by adding servers to either the middleware or the back-end layer.

The drawback of multitier applications is that they are disproportionately complex relative to their functionality. They also can require extraordinary development effort. Microsoft's new DNA 2000 network architecture and COM+ object model address these issues.

Much of the complexity inherent in n-tier design derives from the need to coordinate a task across many servers. Generally, if an activity requires updating two databases--for example, an incoming purchase might require updates on both an inventory database and an accounts-receivable database--both operations must be completed or the entire request must fail. Transaction servers manage this process.

For transactions that don't need to be completed before a client can continue, message-queue servers are sufficient. They enable developers to create transaction applications that are not dependent on the immediate success or failure of access to a database; if one of the databases is not immediately accessible, the transaction can be queued and completed later. However, if the transaction is not completed within a specified time, the entire operation is backed out. This means databases can be in different locations, different organizations, or even different companies, perhaps connected over the Internet, and still have all the advantages of coordinated transaction services.

Message queuing also makes remote or disconnected applications practical. For instance, a salesperson might enter orders on a notebook or CE device, which would then automatically update the database the next time the device connects to the network.

The strong appeal of DNA 2000 and COM+ has less to do with pure technology than with the fact that they extend previous Microsoft standards, which are installed on millions of existing systems. Windows 2000 Server and Microsoft's development tools offer an integrated development environment that allows application development on the ubiquitous Windows NT platform. All the services of a distributed application--including the operating system, middleware services, and even the Web server and database server--come from a single vendor. While there's no need to rely solely on Microsoft for all these services, the dual advantages of unified support and a level of integration not otherwise available carry their own appeal.

continued...page 2, 3