This paper reports the experience of building a highly efficient system for cluster visualization. This implementation was based on the well known master/slave concept which states that, given a "master" node, which is a node chosen to interact with the user and able to process the graphical user interface, splits its display into sections and sends it to "slave" nodes for their individual rendering. Chromium was used to perform the parallel remote rendering operation. In this paper, the detailed aspects of the Visualization Cluster implementation, called MIND, the hardware and software utilized in the project execution and the foremost issues found during the implementation were discussed. A cluster with 16 nodes was employed, connected with a Gigabit LAN. 15 servers, each driving a 20" LCD, were used as slave nodes and one as Master node. Furthermore, the benchmark test results are presented to relate the network bandwidth and the visualization speed of a given data size. An elite design in this system is the development of a script to automate the creation and/or insertion of data into the display node's configuration files using a XML-schema, in so doing, any Linux machine can become a compute node used for the visualization display system.
[1]
Luc Renambot,et al.
TeraVision: a distributed, scalable, high resolution graphics streaming system
,
2004,
2004 IEEE International Conference on Cluster Computing (IEEE Cat. No.04EX935).
[2]
Greg Humphreys,et al.
Chromium: a stream-processing framework for interactive rendering on clusters
,
2002,
SIGGRAPH.
[3]
Malek Adjouadi,et al.
Modeling Web-Based Pediatric MRI Data Repository Site Using OPNET
,
2006
.
[4]
Luc Renambot,et al.
High-Performance Scalable Graphics Architecture for High-Resolution Displays
,
2005
.
[5]
Stanimire Tomov,et al.
Application of
,
2021
.