Blades: innovations and viability for HPC

High performance computing architectures are changing dramatically. Traditional multi-processor, large memory systems may still be the mainstay in the majority of data centers world-wide but new innovations in system design are rivaling these traditional platforms. High performance, high density, cost effective blades are moving into the HPC data center.This presentation reviews the viability of blade technologies for HPC versus more traditional server and cluster technology. The special challenges of the HPC environment are examined. Blades improve data rack consolidation, consolidate the data cabling nightmare and can reduce the overall management effort, but what about power and cooling issues, high performance interconnect availability and processor support and performance? How can the high heat output and power requirements of blades be managed, especially with large configurations? Are the high performance cluster interconnects supported and available for blade cluster configurations? Are blades systems available with high performance, dual core processors?