WebSphere Extended Deployment, Version 6.0.x
             Operating Systems: AIX, HP-UX, Linux, Solaris, Windows, z/OS


Integration of transactional and long-running work

To set up the business grid function to best meet your business objects, it is important to understand how WebSphere Extended Deployment integrates transactional and long-running applications.

For each dynamic cluster you define, WebSphere Extended Deployment determines if the dynamic cluster is used for long-running work or transactional work based on the presence or absence of the runtime environment application, LREE.ear, in the dynamic cluster.

While you can deploy transactional applications to a dynamic cluster meant for long-running work, it is not a best practice. Long-running applications often consume large amounts of memory or CPU, and the transactional application is likely to have poor performance. Conversely, you can deploy a long-running application to a dynamic cluster intended for transactional work. In this case, the absence of the long-running runtime environment on the dynamic cluster prevents the long-running application from receiving any jobs from the long-running scheduler.

The balancer component determines for each node in a node group whether the node runs dynamic clusters that host long-running applications or dynamic clusters that host transactional applications. The balancer decides how to allocate nodes based on the respective service policies that you define for both transactional and long-running applications. If there is more work than the system can handle, then the balancer prioritizes according to service policy importance.

The balancer gives the application placement controller control of nodes that run transactional application dynamic clusters. The application placement controller determines which transactional application dynamic clusters start on which of these nodes. Nodes that are used for long-running dynamic clusters are assigned to the long-running placement logic, which is part of the long-running scheduler. The long-running placement logic determines which long-running dynamic clusters start on these nodes. The balancer can switch a node between long-running and transactional work over time. However, the WebSphere Extended Deployment does not attempt to automatically start both types of work on the node concurrently. If a node is running when the balancer component initializes, then the balancer component attempts to determine how to assign the node based on the dynamic clusters that are running on that component. If no dynamic clusters are started on the node, then the default assignment is sent to the application placement controller transactional work.

Attention: Do not manually start dynamic clusters that do long-running work on nodes that are assigned to do transactional work and vice-versa. This action can cause the system to react in unanticipated ways.



Related concepts
The compute-intensive programming model
Related tasks
Defining a service policy
Concept topic    

Terms of Use | Feedback

Last updated: Nov 30, 2007 4:00:35 PM EST
http://publib.boulder.ibm.com/infocenter/wxdinfo/v6r0/index.jsp?topic=/com.ibm.websphere.xd.doc/info/scheduler/cxdbatchint.html