The SOA Magazine

Whether you are learning about service-oriented architecture or already working with it, you will no doubt have had some anxiety about building a supporting IT infrastructure capable of meeting its runtime demands. Issues such as intermediate layers, large XML payloads, reliability, and scalability across heterogeneous systems have raised concerns among many practitioners. In this article, we’ll cover some of the key technologies that have emerged to address common SOA scalability requirements, such as mid-tier caching, load balancing, and high availability (HA) through service-level grid enablement in support of building out high performance SOA implementations. Collectively, these infrastructure components are referred to as the “SOA grid,” providing enterprise IT professionals with much-needed support for enabling enforceable service-level agreements (SLAs) across entire service portfolios, including Web services, messaging, custom enterprise system applications, and legacy mainframes. Introduction: SOA Performance Challenges Chances are the SOA-based projects you are working on involve building automated processes that combine multiple services into composite applications. These can then be exposed to an unpredictable amount of traffic that originates from external Web users (most likely your customers). As SOA projects become more successful, subsequent projects naturally begin to build upon one another, and new parts of the organization can start realizing the benefits of reuse by tapping into existing services and leveraging composite processes to achieve additional functionality. Even though this can result in an alarming increase in overall traffic both within and outside of the enterprise, the throughput requirements of your business processes cannot be compromised. Automated trading is but one example where even a few milliseconds can make a big difference in the aggregated throughput and performance of the entire system. This section describes different types of technologies that impact this latency as a prelude for the discussion on how an SOA grid can help solve these problems. Scalability, Throughput, and Points of Congestion As you begin to increase traffic in a service-oriented computing environment, the bottlenecks tend to occur in one of three places: • Shared intermediary services Such services perform common integration tasks such as data transformation, content based routing, and filtering. • The services themselves That is, application code exposed as a service and invoked by other services on the network, whether directly or through an orchestration engine. http://www.soamag.com/I10/0907-1.asp (1 of 10)03/09/2007 4:18:34 PM