Implementing Immersive Clustering with VR Juggler

Abstract

The last advances in commodity hardware have allowed users of immersive visualization to create high-performance systems using a set of interconnected computers. These systems, called cluster computers, allow to employ high-quality graphics cards, high-speed processors and significant amounts of memory for much lower costs than would be possible with high-end, shared memory computers traditionally used for virtual reality purposes. In this paper we present ClusterJuggler, a system based on the VR Juggler architecture that enables the use of distributed and clustered computers for the display of immersive virtual environments. We provide and overview of the potential ways to synchronize a cluster for immersive visualization. Then, we describe the ClusterJuggler architecture in detail, and we show how ClusterJuggler allows virtual reality application developers to combine various existing clustering techniques to meet the needs of their specific applications. A performance evaluation of our clustered technique on real 3D real-time immersive applications demonstrates the efficiency of ClusterJuggler with respect to both number of nodes in the cluster and the bandwidth of the interconnection network system.

VR JuggLua: A framework for VR applications combining Lua, OpenSceneGraph, and VR Juggler

“…It has evolved into a commercial solution integrating clustering support and focusing on multi-screen application development. VR Juggler introduced a highly modular architecture for VR applications to provide a "virtual platform" for development and execution on diverse systems [8,7]. Later development extended its use from highend graphics systems to commodity computer clusters [1,7].…”

“…VR Juggler introduced a highly modular architecture for VR applications to provide a "virtual platform" for development and execution on diverse systems [8,7]. Later development extended its use from highend graphics systems to commodity computer clusters [1,7]. The FlowVR platform was developed based on experience in using VR Juggler in a clustered environment, and emphasizes a data-flow model for distributed real-time interactive computation with high modularity [3,2].…”

Analyzing the performance of a cluster-based architecture for immersive visualization systems

“…Since in IV systems the throughput is considered as the effective number of frames per second (denoted as FPS) drawn [37], the time spent for synchronization purposes decreases the system throughput as the number of nodes in the cluster grows. While these constraints are all solved by the hardware of shared memory computers, they must be solved by the software level in a graphics cluster [4]. This is the case of VR Juggler [3,5].…”

“…VR Juggler is a well-known open source suite of software tools that provides a multi-platform framework for the development and execution of VR applications [13,25,27]. Following other approaches [2,26,23,29,33,34,37], the kernel architecture of VR Juggler was modified in order to add clustering support [4] to the features of the development environment.…”

“…This subsystem, called ClusterJuggler [4], enables the use of distributed and clustered computers for recreating immersive virtual environments. The main goals of ClusterJuggler are first, to adapt the cluster software to the particular hardware configuration of the virtual reality systems; second, to provide application portability and scalability from high-end shared memory systems to commodity clusters by hiding the clustering techniques from developers; and third, to allow users to customize their own software methods (based on cluster computing) being used to best meet their specific needs.…”