Building big grids
By DaveLevy on Aug 26, 2008
A colleague of mine, Philipe Trautman presented on winning High Performance Computing deals. He produced some fascinating figures to describe the opportunity. Both Storage and Systems are forecast to grow at double digit rates for the next few years, where as commercial IT is expected to standstill at best. Over 30% of CPUs are going to be bought by HPC solutions during this period and at the moment, 65% of the HPC market is educational and/or research institutes. He outlined Sun's product portfolio consisting of systems, storage, operating systems and interconnects, which can be supplemented by partner products and people. He made the assertion that the real pain is no longer FLOPS, but elsewhere
- Power & Cooling
- Cluster Management
- Application Scalability & Utilization
- Data Access including Filesystem selection
and presumably interconnect architecture and selection. Some of these are problems we have been confronting in commercial data centres for a while, albeit on a smaller scale but the last two are new.
Philipe introduced Dr Wolfgang Hafeman, of "Solutions for Research", a subsidiary of T-Systems and thus Deutsche Telekom, who have built and manage an HPC system for researchers in German commerce and academia, using Sun's products. I wonder if I can get the picture he showed, its quite dramatic. Again, this is an example of the right thing done well. Certainly T-System's people have added massive value to the proposition, although often the success of such a piece of business is based on the quality, drive and determination of the project teams. The relationship between the project teams supercedes the relationship between the companies. Its a great example of partnering for the end customer's success.