Intalio|Works Server is a native BPEL 2.0 process server based on the J2EE architecture and certified for a wide range of hardware platforms, operating systems, application servers, and database servers. Our process server was designed from the ground-up to be deployed on a grid of 1,000 servers by the U.S. Department of Energy (DOE). As a result, it provides the highest level of performance, reliability, and scalability currently available in the industry.

Intalio|Works Server is based on the Apache ODE open source project, which was originally contributed by Intalio.

Intalio|Works Server is a native implementation of the new BPEL 2.0 standard, which added support for distributed transactions and human workflow through standard extensions. This makes the BPEL execution model suitable for large enterprise applications, but significantly more difficult for software vendors to implement. Nevertheless, Intalio|Works Server is backward compatible with earlier versions of the specification, and supports the concurrent deployment of BPEL 1.0, 1.1, and 2.0 processes on the same server. It also supports the seamless migration of BPEL processes from other BPEL engines, such as Oracle BPEL Process Manager and the Sun BPEL Service Engine.

Intalio|Works Server relies on the Ode BPEL Compiler to translate any BPEL processes into "process bytecode" that can be natively executed. This execution model removes the need for custom Java classes generation and deployment, thereby improving performance significantly. This is especially important when a large number of process models must be deployed on a single engine, as was the case for Coghead for example (100,000 process models deployed per server). For runtime process execution, Intalio|Works Server relies on the JACOB framework, which provides support for the persistence of execution state and concurrency.

Intalio|Works Server is a high-performance process engine that can support the most complex business processes, deployed within mission-critical environments. Here is a summary of our performance results, based on multiple actual customer deployments used in production.

• Largest number of individual process activities modeled: 250,000
• Largest number of process models deployed: 100,000
• Largest number of concurrently running process instances: 250,000,000
• Largest number of individual end-users per day: 100,000
• Largest number of servers in clustered deployment: 1,000
• Largest number of in-memory transactions per day and per CPU: 14,300,000
• Largest number of persistent transactions per day and per CPU: 3,600,000
• Roundtrip call to a WSDL web service from a BPEL process: 14 milliseconds