Data deluge from experimental facilities such as the LHC (Linear Hadron Collider) [1] instruments at CERN [2] that generates data at the rate of 2GB/s to complex simulations and observational data pose visualization challenge to support real-time high resolution data analysis. The National Grid Service (NGS) [3] and EGEE [4] provide distributed, high performance storage and compute resources to support research communities for data management, simulation runs and data processing but not visualization and interactive data analysis. In general, until recently proprietary high performance graphics computers were used to analyse very large datasets in a post processing stage. Increasingly, high performance graphics cards, processors and high speed interconnect are used together to build graphics clusters similar to compute clusters. These generic systems are affordable and support open source software stack.
The vizNET STFC partners have installed a 17 node, 64-bit commodity cluster with dual processors on each node with Linux operating system. We installed a variety of visualization and other applications and test suites, met conformance testing and have gained NGS [3] affiliation as an associate resource. This system is made available to the academic community under a joint STFC/vizNET/NGS banner.
Chromium [5] and VirtualGL [6] are the two software packages that are used for this purpose. Understanding the architecture these enhances the ability to use the system effectively. This document sets out the details of these open source applications that can be used to access our parallel visualisation services on the cluster from remote desktops, describes the hardware and how it can be used.
Download the full report here