Distributed Multithreaded Volume Rendering Utilizing Dynamic Scheduling

Joseph Hennessey, JHU

Volume rendering is an important visualization technique for medical researchers but is computationally expensive. Most labs have multiple computers available but no one computer may have the ability to render at sufficient speeds. Towards this end I have developed an algorithm and implementation for distributing volume rendering over a network of heterogeneous computers that efficiently utilizes the available computational resources. Additionally the system dynamically adapts to distributed user load so as to minimize the performance impact on local users of the remote systems.