Tuesday Jun 02, 2009

Europe's largest supercomputer

The Forshung Julich phase two super computer, now Europe's largest, had its formal opening session last week and Mark Hamiltion, Sun VP who leads our HPC team went to visit them, and recorded it on his blog, in a couple of articles dated as at the end of May, because it runs on Sun. He wrote three articles, several of them with lots of pictures.

This is Europe's largest super computer and runs on Sun's Constellation systems, Mark's article "Memorial Day in Germany" and the Forshung Julich web page, "Systems Configuration" talk about the technology, where they state, they have 2208 compute nodes, each with dual, Intel Xeon X5570 (Nehalem-EP) quad-core processors, running at 2.93 GHz. This has over 17500 cores with 207 Teraflops peak performance, hardly surprisingly they have also taken four of Sun's Data Centre Switches.

The EU's PRACE project funded the feasibility of this and I have been tracking it for a while since we knew that phase 2 was to be based on Sun's hardware. I have a link roll...

<script TYPE="text/javascript" SRC="http://feeds.delicious.com/v2/js/DaveLevy/julich?title=My%20Delicious%20Bookmarks&icon=s&count=7&sort=date&tags&extended"></script>

which records a bunch of pages about it and this page, the Juropa Supercomputer has a rather cool picture.

the JuRoPA Super Computer

which I have linked to, but shrunk to get on this page.

tags:

Tuesday Aug 26, 2008

Building big grids

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
  • Space
  • Cluster Management
  • Consolidation
  • 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.

tags:

About

DaveLevy

Search

Archives
« April 2014
MonTueWedThuFriSatSun
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
    
       
Today