My last post was motivated by the awareness of the opening of a opportunity window in the HPC / supercomputing field and therefore a kind of apology for research in this field in Spain or UE.
It is not that UE is not aware about the fact that the availability of supercomputing infrastructure is a key component of scientific and industrial advance and competitive advantage. As a proof the PRACE project. This post pretends to be a very short summary of this project.
As any agent involved with in science and technology activities knows, the Framework Programs (FP) are the main EU financing tool in R&D. By now we are living in the FP7 quinquennial program, lasting from 2008 to 2013. In the EU web site Europa, in the Cordis area one can find a lot of information about EU R&D activities and the FP7.
One of the projects included in FP7 is the PRACE project. Supercomputing is considered as an infrastructure (and therefore a transversal technology) and the aim of the project is to build a pan-European Research infrastructure (RI): in short a network of HPC systems where each Tier-0 system will be a supercomputer.
As any corporation in this sector Bull has a quite rugged history. From wikipedia: “Bull was nationalised in 1982 and was merged with most of the rest of the French computer industry. In 1994 the company was re-privatised”. Now 70% of its capital floats on stock markets and not surprisingly 43% of its revenues is from the Public Sector.
On the other hand I have not been surprised to learn that the third Tier-0 system is not from EU. I plan to post some day about the global industrial organization of the HPC sector. There may be more EU players in this sector than I think. But this post was supposed to be about PRACE project, so enough comments about the HPC industry players !
The best part of the PRACE information brochure is when they describe the 22 scientific applications. Nice, and possibly just the tip of the iceberg. And now some surely uninformed and unfair criticism: if the plan is to import from abroad HPC manufacturers the Tier-0 elements of the PRACE network (as it seems to be the case), I would say that this looks like a missing chance to refresh or revive the UE-HPC industry with the newly open opportunity window we commented in previous post. Maybe this can be corrected in FP8 (if not too late); I do not know how rigid are these FP regarding program and project´s initial conception or maybe the sentence “Assessed hardware and software components for future multi-Peta flop/s systems on nine prototype systems” in the brochure already addresses to this issue. Probably most of the future Tier-0 elements already available (as the spanish Marenostrum) will be outdated soon, and could be replaced by local manufacturer products, if competitive of course. In any case, I have to study in deep this project.
For closing this post I could not resist to link to a great blog entry I have found today, and apparently unrelated with the rest of the post content, about how Facebook has been tailored. What a complex software system this social network platform is ! And how useful supercomputing surely is for this kind of Web 2.0 services ! Using the words of Facebook Vice President of Engineering quoted in the article:
“Scaling any website is a challenge,” Schroepfer said, “but scaling a social network has unique challenges.”
He went on to say that unlike other websites, you can’t just add more servers to solve the problem because of Facebook’s “huge interconnected dataset.” New connections are created all the time due to user activity”.
Besides PRACE, EU has stablished is own project or initiative (starting in June 2010) for exascale supercomputing, the EESI. According to a report from this public agency, it is confirmed that Bull is the unique european vendor of supercomputing systems.