Theme and Laboratory	Outline
Parallel Protein Information Analysis System and Parallel No-Cutoff Molecular Dynamics Simulation on a Compact 8-node Linux PC-Cluster Parallel Application TRC Laboratory http://pdap1.trc.rwcp.or.jp/research/research.html	We have developed biological and chemical parallel applications running on our PC clusters, such as the PAPIA system and the MolTreC no-cutoff molecular dynamics simulator. We have successfully ported those parallel application programs onto a new compact 8-node Linux PC cluster which we will demonstrate at the digital poster session of the RWC 2000 symposium. The PAPIA (PArallel Protein Information Analysis) system is an integrated parallel computing system dedicated to typical search/ analysis problems required in protein research, such as structure similarity search, sequence homology search, and multiple sequence alignment. The system was first implemented on a 64-node PC cluster (PAPIA cluster) and we have been operating a WWW-based free service since April 1998. This service has been accessed from 63 countries in the world to date. Now the PAPIA system also runs on a compact and inexpensive 8-node cluster (PAPIA mini cluster). We also present our work on chemical applications with our parallel molecular dynamics simulation. We have developed an original parallel molecular dynamics simulation program called MolTreC which employs a parallel Barnes-Hut tree code for calculating long-range Coulomb interactions. The MolTreC program can rapidly calculate the Coulomb interactions of proteins in water without any cutoff approximations. The PC clusters show good scalability in these two real applications.
Agent based Middleware System Applied to Particle-Mesh Heterogeneous Scientific Simulations Parallel Application Hitachi Laboratory http://www.rwcp.or.jp/activities/acheivements/PA/hitachi/hp1new.html	A new middleware system is presented that provides high-level transparency in the form of multiple agents to automatically transfer physical values and transform the length and time scales between different kinds of scientific simulations. Using a correlative specification between different discrete points such as those found in mesh and particle methods, the agent system is designed to provide semantic transformations as well as appropriate communication paradigms to effectively combine extensive scientific simulations in the unified way. Several kinds of particle and mesh heterogeneous simulations are shown to demonstrate the general applicability of our system. These combinations can be realized using agent library calls which amount to less than 1/10 of the programming steps when compared to usual development techniques.