An Extensible Scientific Computing Resources Integration Framework Based on Grid Service

Scientific computing resources (e.g., components, dynamic linkable libraries, etc) are very valuable assets for the scientific research. However, due to historical reasons, most computing resources can't be shared by other people. The emergence of Grid computing provides a turning point to solve this problem. The legacy applications can be abstracted and encapsulated into Grid service, and they may be found and invoked on the Web using SOAP messages. The Grid service is loosely coupled with the external JAR or DLL, which builds a bridge from users to computing resources. We defined an XML schema to describe the functions and interfaces of the applications. This information can be acquired by users by invoking the "getCapabilities" operation of the Grid service. We also proposed the concept of class pool to eliminate the memory leaks when invoking the external jars using reflection. The experiment shows that the class pool not only avoids the PermGen space waste and Tomcat server exception, but also significantly improves the application speed. The integration framework has been implemented successfully in a real project.