Should the Nordic countries continue to use PRACE?

27.02.2014
The Partnership for Advanced Computing in Europe (PRACE) provides Europe’s most extensive high-performance computing systems – 100 000 times more powerful than an ordinary PC. As PRACE is considering changing its payment model, the Nordic e-infrastructure Collaboration (NeIC) decided to evaluate Nordic researchers’ use of the resources.

Today researchers are using PRACE resources for free, competing with each other to be awarded access based on a peer-review procedure. It is likely that PRACE will introduce a pay-for use policy. So what does Nordic research actually gain from using PRACE resources? Will it be worth paying for access or will it be more cost-efficient to create a Nordic facility? To get the facts straight, the NeIC commissioned an independent evaluation of the Nordic use of PRACE. The evaluation committee was chaired by Dr Rob Pennington, CTO and Deputy Director, National Center for Supercomputing Applications (NCSA), University of Illinois at Urbana-Champaign. Dr Pennington came to Copenhagen to present the findings of the evaluation report on 12 February 2014.

 

PRACE enables scientific success

The evaluation committee estimates that the Nordic region’s need for PRACE capabilities corresponds to roughly 25 per cent of a Tier-0 system, in addition to current national capabilities. The committee has drawn up seven recommendations, divided into organisational, scientific and technical recommendations.

 Dr Rob Pennington

Dr Rob Pennington sums up the committee’s findings:

- I think PRACE has been very scientifically successful in providing access to resources for scientists who have problems that are too big, or too risky to be funded in their national systems. At PRACE, researchers have a very limited amount of time. Then they have to re-compete against other world class researchers, and this pushes the limits and can give way to the unexpected.

How much use do the Nordic countries make of PRACE?

- The number of projects and amount of time projects led by Nordic PI’s are getting is impressively high. The Nordic researchers are competing successfully. I think it is because they have good infrastructure within their countries, they’ve got very good education, and they have the incentive, the desire and the opportunity to build up within their country and then go on to a larger scale.

Dr Pennington continues:

- In the Nordic region, there are some very active groups of researchers and some very capable programmes. The possibilities for extending that cooperation are very interesting. And I think the Nordic region has got enough of an identity as a region, and this is a good start. You have different national identities, but the communications and collaborations are certainly in place.

 

Regarding the transparency of the PRACE organisation, the evaluation committee is more hesitant:

- We have learned that getting information from PRACE on projects that were not awarded for example, or on how many people have applied, is difficult. It is a delicate balance between respecting the researcher’s need for privacy and the need for transparency. But it seemed that it was unnecessarily difficult for us to get the information we were looking for, Dr Pennington explains.

 

Everything comes back to data

High-performance computing aggregates computing power into systems that are many orders of magnitude more powerful than ordinary PCs. High-performance computing systems can be used to solve the Grand Challenges that would otherwise be impossible to solve. Application areas include climate science, life sciences and physics, among many others. Dr Pennington discusses the reach of e-Infrastructure into our day-to-day lives:

- E-Infrastructure is interesting, everything comes back to data and information and how you process it. How do you make your priorities and how do you understand what is actually happening? How do people interact? How can you provide information so that you make better informed choices? Look at the changing population demographics – how do you build the infrastructure for the political, economic and social environment to support our children and the next generation after that? These are hard questions but we are finally at a point where we can begin to understand how groups of people react. This is different from computing and physics where a lot of this started, where we can solve the equations. In societal questions, we are not even at the point of really understanding the questions to the point of being right down to the definition of what the problem is.

 

Next Nordic steps in PRACE

The next pressing question is how the NeIC will respond to the evaluation. Director of the NeIC Gudmund Høst comments on this:

- From the evaluation we have learned that Nordic researchers make use of roughly 25 per cent of a Tier-0 resource. Thus, an alternative to PRACE participation for the Nordics in the future could be to invest in a moderate Tier-0 facility. If a future payment model for PRACE is based on national GDP it could prove costly for the Nordic countries to participate, and an alternative Nordic Tier-0 facility might be more cost-efficient. The evaluation committee also had a hard time extracting the relevant information for their work. The implied lack of transparency is of serious concern because it may hamper the optimal use of the national-Nordic-European ecosystem of resources.

The NeIC will discuss the evaluation report at their June meeting.

Gudmund Høst

 

What do you find most important in the evaluation report?

- Nordic research groups have a high success rate on their applications for PRACE resources. This is very positive, and it shows that we have excellent research communities that can make good use of leading edge resources to the benefit of scientific progress.

Dr Høst continues:

- I’m also glad to learn that the committee concludes that the NeIC is a well-suited instrument for coordinating and cooperating on high-performance computing. Furthermore, the committee recommends that a future strategy for Nordic collaboration on high-performance computing is based on future research challenges. An important next step would be to build a scientific case through translation of researcher needs into technical requirements.

 

Read the evaluation of the Nordic Participation in PRACE

 

Members of the evaluation committee:

CTO Rob Pennington, the National Center for Supercomputing Applications (NCSA), University of Illinois (Chair)

Distinguished Professor of Computer Science and Informatics Geoffrey Fox, Indiana University

Executive Director Alison Kennedy, EPCC

Visiting Professor Ron Perrott, University of Oxford

 

Recommendations of the evaluation committee:

Organisational:

1. Put in place an effective framework for Nordic cooperation and coordination.

2. Maintain open and transparent processes at all scientific, technical and organisational levels.

Scientific:

3. Develop the scientific goals, justifications and use cases.

4. Engage nationally and internationally in projects to meet the scientific goals.

Technical:

5. Determine the ranges of user requirements.

6. Develop the technical roadmaps for Tier-0, Tier-1 and other national systems.

7. Develop the technical expertise to support scientific success.

 

Text: Linn Hoff Jensen

Newsletter
Facebook