On 11/12/07, Sun announced new systems designed to address the extreme computation, scale, and storage requirements of high-performance computing (HPC) customers. Sun Vice President of Global Systems Engineering, Hal Stern, invites Solaris Engineering's distinguished engineer, Josh Simons, to this latest edition of Innovating@Sun to double click on the benefits and rationale behind these innovative new technologies. Stern and Simons discuss:
Sun's Constellation System, the world's first open HPC architecture capable of scaling from departmental clusters all the way up to the largest supercomputer in the world.
A set of components that allow customers to build petascale systems at the high end of the HPC market and, taking those same capabilities, scale them downwards.
Support for SPARC, Intel, and AMD processors
A shift from big iron to a clustered approach
Infiniband as the technology of choice for high-bandwidth, low-latency solutions
Addressing customer pain points at large scale including power, switch complexity, cable complexity, fatter nodes, and driving down node count
How this design point of modular computing becomes a prototype for building future datacenters
Appealing to a wide variety of developers
Getting software to scale to appropriate levels for system management, programming, etc.
The importance of storage in HPC
Why Sun is hitting HPC with a vengeance
The rationale of pushing Solaris if the vast majority of the HPC market is Linux-focused.
Sun Constellation System