Capacity and Capability Computing using Legion
190 likes | 317 Vues
The Legion Project at the University of Virginia focuses on Capacity and Capability Computing to enable extensive computational experiments across heterogeneous resources. Capacity Computing allows for larger experiments by utilizing more resources for a single problem or multiple independent problems. Capability Computing introduces new methodologies for tackling diverse challenges, fostering collaboration across various domains. With features supporting fault tolerance, security, and the integration of legacy systems, Legion provides a robust framework for applications in biochemistry, climate modeling, and more.
Capacity and Capability Computing using Legion
E N D
Presentation Transcript
Capacity and Capability Computing using Legion Anand Natrajan (anand@virginia.edu) The Legion Project, University of Virginia (http://legion.virginia.edu)
Capacity and Capability Computing • Capacity Computing: Conduct larger computational experiments by expending more resources • single problem • multiple, independent problems • Capability Computing: Conduct experiments with new mechanisms • Heterogeneity • Security • Collaboration
Grid Environment • Disjoint file systems • Disjoint namespaces • Multiple administration domains • Unpredictable load, availability, failures • Security problems • Computers • Networks • People • Data • Devices
Wide-area High Performance Complexity Management Extensibility Security Site Autonomy Input / Output Heterogeneity Fault-tolerance Scalability Simplicity Single Namespace Resource Management Platform Independence Multi-language Legacy Support Grid OS Requirements
MPI / PVM P-space studies - multi-run Parallel C++ Parallel object-based Fortran CORBA binding Object migration Accounting Remote builds and compilations Fault-tolerant MPI libraries Post-mortemdebugger Console objects Parallel 2D file objects Collections Licence support Tools
Protein Folding with CHARMM Molecular Dynamics Simulations 100-200 structures to sample (r,Rgyr ) space r Rgyr
Resources Available IBM SP3 UMich 375MHz Power3 24/24 HP SuperDome CalTech 440 MHz PA-8700 128/128 DEC Alpha UVa 533MHz EV56 32/128 IBM Blue Horizon SDSC 375MHz Power3 512/1184 Sun HPC 10000 SDSC 400MHz SMP 32/64 IBM Azure UTexas 160MHz Power2 32/64
Transparent Remote Execution • User initiates “run” • User/Legion selects site • Legion copies binaries • Legion copies input files • Legion starts job(s) • Legion monitors progress • Legion copies output files
Mechanics of CHARMM Runs Register binaries Legion Create task directories & specification Dispatch runs Dispatch more runs
Types Of Applications • Legacy applications • Legion-aware applications • I/O library • 2D file object • Applications Using Stdgrid • Parameter Space Studies • Parallel Programs • MPI, PVM, MPL, Basic Fortran Support (BFS)
Computing in the Near Future • Security • Fault-tolerance • Heterogeneity • Collaboration • … • Legion supports these and other needs
Heterogeneous Runs BT-Med Ocean Model
Cross-Organisation Collaboration • Different companies • Proprietary simulations and data • Each needs the other • Form virtual partnership
Flexible Context Space Disk e ftp Directory NFS HTTP Samba FTP Context Context Context legion_export_dir legion_import_tree Disk Directory Context Directory Directory
Interfaces • Samba, NFS, FTP, HTTP interfaces to distributed file system • Windows interface for file sharing • Command-line through Unix-like tools • Web interface through browser • Programmatic interfaces through system calls in C, C++, Fortran, Java
Windows NT, 2K, 98, 95 Sun (Solaris) SGI (Irix, Origin) Intel (Linux, Free BSD) DEC (Unix, Linux) Cray (T90, T3E) IBM (AIX, SP-2) HP (HPUX) Nimrod Codine LoadLeveler Maui PBS NQS LSF Platforms
Applications • Biochemistry and Molecular Science • Information Retrieval • Materials Science • Climate Modelling • Neuroscience • Aerospace • Astronomy • Graphics NPACI - SDSC, UCSD, Caltech, UTexas, Umich, UCB, UVa. DoD MSRCs - NAVO & ARL, NASA Ames