Outline
Background - Distributed computing at DL
Commodity-based and High-end Systems
- Single-node performance & the Interconnect bottleneck
- Prototype Commodity Systems; CS1 - CS20
- High-end Systems: IBM SP/p690+, SGI Altix 3700, plus HP/Compaq Alpha Server SC and SGI Origin 3800
- Performance Metrics
Communication benchmarks
- GbitEther, Myrinet, SCI, Infiniband, Quadrics ..
- MPI and Global Arrays (GA tools)
- Pallas PMB MPI Benchmark - point-to-point and MPI Collectives (V 2.2 - 16, 64 and 128 CPU systems) -
- Pallas Bandwidth Benchmark - B_EFF (EFF_BW)
- Global Array benchmarks (V 3.4)
- NWChem, GAMESS-UK and MOLPRO