HP Encina/9000: middleware for constructing transaction processing applications - includes glossary - Product Information

Hewlett-Packard Journal, Dec, 1995 by Pankaj Gupta

A transaction processing monitor for distributed transaction processing applications maintains the ACID (automicity, consistency, isolation, and durability) properties of the transactions and provides recovery facilities for aborting transactions and recovering from system or network failures.

Transaction processing systems are widely used by enterprises to support mission-critical applications, such as airline reservation systems and banking applications. These applications need to store and update data reliably, provide concurrent access to the data by hundreds or thousands of users, and maintain the reliability of the data in the presence of failures.

The HP Encina/9000 transaction processing monitor[1] provides the middleware for running transaction processing applications. It maintains ACID (atomicity, consistency, isolation, and durability) properties for transactions (see the glossary on page 65). It ensures that applications that run concurrently will maintain data consistency. Encina/9000 also provides recovery facilities for aborting transactions and recovering *om failures such as machine or network crashes.

DCE and Distribution

Encina/9000 provides the ability to write distributed applications. Encina/9000 applications can be written as clienVserver applications with the client and server possibly running on different machines. Encina/9000 servers can communicate and cooperate with each other in updating data on several different machines.

Distributed applications provide several advantages. The data maintained by an enterprise may itself be distributed because of historical and geographical considerations. Furthermore, distributed applications are able to exploit parallelism by running concurrently on several machines.

Distributed computing offers the advantage of improved performance, availability, and access to distributed data. Performance is improved by spreading the computing among various machines. For example, the application's user interface can be run on a PC while the user code could be split to run on several machines. The use of multiprocessing machines to provide parallelism for multiple users can improve the throughput of the system. Availability can be increased by a distributed system in which replication is used to keep several copies of the data. Access to distributed data or to data that is maintained in several databases is also facilitated by distributed computing.

Data may be distributed because the database becomes too large or the CPU on the database machine becomes a bottleneck. Data can also be distributed to increase availability and improve the response time by keeping the data close to the users accessing it. Finally, data can be distributed to keep separate administrative domains, such as different divisions in a corporation that want to keep their data local.

Encina/9000 uses the Open Software Foundation s DCE[2] (Distributed Computing Environment) as the underlying mechanism for providing distribution. It uses the DCE RPC mechanism to provide client/server communication. Encina/ 9000 is also very closely tied to DCE naming and security services (see the articles on pages 28 and 41 for more about these services). For example, an Encina/9000 server can be protected from unauthorized use by defining access control lists (ACLs). ACLs contain an encoding of the authorization policy for different users and are enforced by DCE at run time. ACLs are described on page 49. Encina/9000 also makes use of the threading package provided by DCE.

To achieve optimum price and performance, careful consideration needs to be given to how the data and the application are partitioned. Throughput and response times are often the key criteria by which users judge the performance of a system. Encina/9000 provides the flexibility of being able to specify the distribution topology of the application. In addition, users can specify data replication if it will help to ensure higher availability of mission-critical data.

Two-Tiered versus Three-Tiered Architectures

In the past, transaction processing applications were implemented using a two-tiered architecture (see Fig. 1). In this paradigm an application is written as a client, which accesses a database server. The client implements the graphical user interface (GUI) and the application logic. The database server handles access to the data stored in a database.

The advantage of this approach is simplicity. The disadvantage is that it is not scalable beyond a certain point. It is also less flexible and harder to modify to meet new business needs.

Encina/9000 allows the development of applications using a three-tiered architecture like the one shown in Fig. 2. In this paradigm, an application is partitioned into a client that implements the graphical user interface to the user, an application server that implements the business logic of the application, and a database server that implements the database access.

The Enc na/9000 three-tiered architecture provides the following advantages over traditional two-tiered architectures: