Network Computingwww.networkcomputing.com

To evaluate these middleware products, we developed small test applications in Java, then gauged the ease with which we could incorporate each middleware product into our test environment. In each class, we wanted to find the product that presented the fewest limitations, was easiest to configure, offered the most security and performed the fastest. We also applied category-specific criteria, such as database neutrality in the JDBC drivers.

WebLogic's Tengah was the Swiss Army knife of Java middleware in our evaluation. The comprehensive product garners our Editor's Choice award with its useful implementations of functions in several middleware categories, including JTS and JDBC. Indeed, only Tengah offered JRMI and JNDI services in the lab. Our hopes that a Java-based transaction monitor product would give us JRMI and JNDI were dashed.

A few products in individual categories stood out for their versatility and excellent performance. OpenLink Software's JDBC drivers were small and fast, IBM's MQSeries handled heterogeneous platforms with ease, and the combination of BEA Systems' TUXEDO and Jolt made for a potent DTP tool.

Java Application Server
WebLogic Tengah 3.0
Giving us a raft of JDBC, JRMI, JTS and JNDI services in a single, well-designed package, Tengah impressed us with its overall support for several Java middleware functions. We particularly liked Tengah's multithreaded approach to providing Java middleware services, its transaction load-balancing and its dynamic application partitioning feature. The graphical management console was icing on the cake, letting us monitor and configure multiple Tengah-based applications from a central site. Tengah stood out among the other Java middleware products as a cohesive, consistent and comprehensive collection of plumbing services.

Tengah's multithreaded architecture pooled resources, cached database query results and shared database connections among Java applications. As we increased the workload and added more Tengah servers to the network, the product scaled quite well. It surprised us by exhibiting good performance, an attribute Java software often lacks. WebLogic claims Tengah has multiserver clustering support for reliability and fault tolerance. Although we didn't test clustering, we did note Tengah replicated session and state information across servers.

At our behest, Tengah used SSL (Secure Sockets Layer) encryption to shield our network's messages from curious eyes. We also configured it to authenticate network traffic via digital certificates that we issued through Microsoft Certificate Server.

During our tests, we moved application components to different servers to evaluate Tengah's JRMI abilities and its JNDI location transparency feature. In each test, Tengah published the location of our Java test application objects via JNDI and successfully invoked them no matter where on the TCP/IP network they resided. Using the product's dynamic application-partitioning feature, we relocated components without halting the application. Tengah distributed transactions across application servers without interruption and let us divide the transaction workload among the various business logic components on those servers.

We used Tengah's event manager to publish events on the network from within our Java applications. In the tests, we instructed our business logic on the application server to wait for notification from a Java applet client that the applet had finished collecting information and had stored that information in the database. Tengah announced each event to that event's subscribers, triggering the business logic's retrieval of database entries and calculation of results. Each subscriber registered interest in the particular event, by type of event.

WebLogic ships a multitier JDBC driver class with Tengah. We easily accessed Oracle, SQL Server and Adaptive Server databases via Tengah's JDBC drivers.