In distributed shared memory cluster environment, the dynamic load balancing is a very critical issue. Various techniques are used in order to distribute the load dynamically among different nodes. A good balancing scheme needs to evenly distribute the workload among the available processors and locate the tasks close to their data so as to reduce the communication and execution time. In this work, we investigate "work stealing" on distributed shared memory clusters. We propose a dynamic load balancing model with "work stealing" which intelligently balances the load among different nodes resulting in efficient use of system. The "work stealing" typically completes the tasks more than twice as quickly, despite being allotted the same or fewer processors. "Work stealing" consistently provides higher utilization when many jobs with varying characteristics are using the same distributed shared memory cluster system.