As with any parallel and distributed data processing system, load balancing is an important issue. For best possible system throughput and response time for user interactions, resource utilization has to be optimized. In this context we are developing a dynamic load balancing interface and integrate it with Equalizer's ability to distribute the workload among available graphics resources. The targeted solution will dynamically assess the workload and resource availability and assign the most optimum work distribution to the parallel resources considering parameters like resource power, locality of data as well as communication and synchronization requirements and dependencies. Apart from a feedback loop at runtime, the load balancing infrastructure will expose an API to the application for gathering timing and statistics data, and distribute the work accordingly. In order to improve the accuracy of load prediction, we employ a feedback control algorithm from control theory called PID controller.