The really interesting bit is that the multiple server types they offer are differentiated by more than the traditional two axes - CPU & memory. They offer hybrid nodes with GPU or Tilera accelerators.
Presumably they are either solving some thorny user-multiplexing issues, or are resolving whole machine instances for each user. Tough choice.
1 comment:
Oh, and where are the FPGA options? It's not like SGI is a stranger to FPGAs
Post a Comment