A Workstation Cluster for High Performance Computing in Structural Biology

点击翻译按钮获取中文摘要

The objective of this proposal is to continue the development of a network of high performance workstations coupled through an Asynchronous Transfer Mode (ATM) switch for effective utilization in computational structural biology. The funds requested will complement (matching) funds to be obtained from Hewlett-Packard Comp. to build a cluster of sixteen (8 to be funded through NSF) Hewlett-Packard workstations based on PA-RISC 7200 processors, to acquire a communication backbone for this cluster (based either on an ATM or on fibre channel technology, which will be decided at the time of purchase), and to upgrade the ATM switch of an existing cluster of 16 workstations. The two clusters will be used: (1) for ongoing collaborative projects with experimental laboratories which involve very large scale modeling of supramolecular systems (membranes, phopspho lipase A2 - membrane complexes, polymerization of G-actin to F-actin and association of gene V protein with single stranded DNA, complexes of nuclear receptor proteins with DNA, light harvesting systems, and others); (2) for NMR protein structure/function studies; (3) for studies of electron transfer paths in redox proteins and for the prediction of protein structures (protein folding). One of the clusters will allow users to pool small numbers of workstations (1-4), the other will allow a pooling of 8-16 workstations. The peak power of a 16 workstation cluster will also be utilized for real time modeling. For effective parallel computing across the clusters molecular dynamics programs (namd, vmd) have been developed and are being further improved. The effort is funded through an NIH Resource grant and through an NSF Grand Challenge Applications Group grant. Real time modeling will be based on a new graphics program (vmd) for large screen stereo projections and 3D interactive devices. Besides the three principle investigators, sixteen experimental investigators will benefit from the computational resources sought in this proposal. Two computer science faculty members collaborate with us on molecular dynamics algorithms and parallel computing software. The computational work is partially funded through a grant from the Carver Charitable Trust which has provided six graduate students and two postdoctoral positions for projects between the computational group operating the clusters and experimental laboratories.