Efficient load balancing for cluster-based parallel real-time rendering

• Research unit: INNOSUISSE
• Project number: 9394.2;8 PFES-ES

Short description (German)

Efficient load balancing for cluster-based parallel real-time rendering

Short description (English)

Efficient load balancing for cluster-based parallel real-time rendering

Abstract (German)

This project addresses the development of efficient load balancing methods and implementations for highperformance cluster-parallel visualization, based on real-time rendering using OpenGL. The project includes two major objectives: The first is the development of a flexible load-balancing mechanism and its integration into the existing parallel rendering framework Equalizer. The second is the development of an analytical cost model that describes the expected parallel rendering performance given the data set, rendering and display parameters. From the cost model an efficient heuristic will be derived and combined with the load balancer in Equalizer resulting in an advanced and commercially competitive industrial software solution.

Abstract (English)

This project addresses the development of efficient load balancing methods and implementations for highperformance cluster-parallel visualization, based on real-time rendering using OpenGL. The project includes two major objectives: The first is the development of a flexible load-balancing mechanism and its integration into the existing parallel rendering framework Equalizer. The second is the development of an analytical cost model that describes the expected parallel rendering performance given the data set, rendering and display parameters. From the cost model an efficient heuristic will be derived and combined with the load balancer in Equalizer resulting in an advanced and commercially competitive industrial software solution.