Placing a price-tag on e-infrastructure

What is the cost, impact or value of e-infrastructures? A number of projects are looking for answers to these not-so-simple questions, so funders can make important decisions about the future investments of scientific research. Image courtesy jitheshvv on stock.xchng.

The recent growth of services from commercial cloud computing, such as Amazon’s EC2 and Microsoft’s Azure, has provided science with a cost-per-use model for research.

“Traditional approaches are being challenged by the development of grids and clouds, and opportunistic computing models, which offer alternatives, that were not imagined ten years ago. It is important for those of us who fund e-infrastructures to examine these new models to determine where they provide intellectual and economic advantages, and this will lead to scientific progress in the 21st century,” said Barry Schneider, program director for the eXtreme Science Engineering Discovery Environment (XSEDE) at the National Science Foundation in the US.

Public-funded e-science in Europe is trying to find the real cost of running e-infrastructure. “Fundamentally, it comes down to different business models. Cloud providers like Amazon or Microsoft always have to provide a clear price to customers, so they also need to have a clear internal understanding of costs,” said Owen Appleton, director at Emergence Tech Limited, UK.

One cost source is the European Grid Infrastructure - Integrated Sustainable Pan‐European Infrastructure for Researchers in Europe (EGI-InSPIRE), a European Commission funded project, that manages the largest e-infrastructure in Europe. It has a project cost of €72 million ($96 million) over four years, which is funded partly by the European Commission, and partly by national grid initiatives. 

Multinational e-infrastructures are a much bigger task.

Big questions require bigger surveys

The scale of the question becomes difficult when you have to look at multinational e-infrastructure partnerships in Europe, such as the European Grid Infrastructure, which is funded through a mix of local, national, international, and other funding sources. Europe has hundreds of computing centers, with tens or hundreds of funding sources each.

“As e-infrastructures generally don't charge for use in the same way, they control their costs on a different level. Who paid for which fraction of each CPU hour is incredibly hard to determine,” said Appleton.

This map highlights the number of national grid initiatives (in green), supercomputer centers (in blue) and national research & education networks (in red) in Europe. The e·nventory project is helping to create simple online tools for researchers and funders to analyze, assess, and disseminate the impact of e-infrastructures. Image courtesy Jorge Sanchez.

“The crucial factor, however, that made it difficult in the past to estimate costs, was the lack of coordination entities at a European level,” said Jorge Sanchez, coordinator of the e·nventory project, which is assessing the impact of e-infrastructures.

This is one of a number of projects currently trying to establish a summarized cost or value to multi-national e-infrastructure. “All in all, although difficult, it is not impossible to collect cost if a systematic process is put in place,” said Sanchez.  

Appleton is part of the e-FISCAL project, which is trying to evaluate the cost of running European grid and high-performance computing [HPC] resources with a survey that was released last month.

“On the very basic level, the e-FISCAL survey produces European averages for different cost categories, which can be used as a tool to identify areas that are good candidates for optimization,” said Appleton.

E-FISCAL’s results will come out this summer. “A response can cover a single center, or represent a national grid infrastructure, the number of responses we need might be anywhere from 15 to 150. We will answer questions like: how much money would I need to run the system for the next year if I was in charge of the whole thing?”Appleton said.

But, cost is only part of understanding impact or value.

Assessing the true value

For every one pound sterling (1.6 dollars) of UK public spending in e-infrastructure, 10 pounds sterling (16 dollars) of net Gross Added Value (GVA) will be generated within two years, rising to 25 pounds sterling (40 dollars) after five years, according to the 2011, A Strategic Vision for UK e-Infrastructure report.

The e-FISCAL survey is looking at the estimated operational costs from e-infrastructure centers to make data collection easy enough so they can ask participants to gather it. Then, they can simulate annual infrastructure costs for 2010 and 2011 to give an estimated annual cost of ownership for those two years. Image courtesy Sandra Cohen, of the Athens University of Economics and Business, Greece.

This suggests e-infrastructures are valuable investments; but, “cost evaluation, is just one piece of the story. We should think about the impact of e-infrastructures,” said Andrea Manieri, project coordinator for the ERINA+ study, a European-funded project that is looking at the socio-economic impact of research e-infrastructures.  

“Impact is about costs versus benefits. Higher costs may be fully justified by extra benefits. During this period of economic crisis the government and public authorities, either at national or European level, need sound instruments to qualify the expected benefits from each investment. Spending fewer Euros is not always the best option: more important is to be sure that for each Euro invested the expected return could be greater,” said Manieri.

The ERINA study found that researchers are unfamiliar with doing an impact assessment methodology for their own results or technologies. To date, projects typically use qualitative rather than quantitative or unambiguous information to assess themselves. The ERINA+ project aims to close this gap said Manieri.

ERINA+ proposes a new methodology to assess the impact of European Commission funded e-infrastructures and to provide tools for researchers to independently perform an impact assessment of their projects.

Manieri said, “this will be achieved through a wide outreach strategy including focus groups, workshops and a final conference, leading to the production of a final set of recommendations published as a white paper. ERINA+ will give tools to about 20 projects as a trial in February 2012.”

Meanwhile, e·nventory, a co-funded project by the European Commission, has built an online platform called the European e-Infrastructures Observatory, with interactive and user-friendly visualizations, to give a clearer picture to funders, researchers, and the public, of the impact and progress of European e-infrastructures.

The number of CPU hours that national grid initiatives are using is increasing rapidly. Tools that help assess the impact of e-infrastructure research are essential to understanding the value they make to society. Image courtesy Jorge Sanchez.

“With a set of seven visualization tools, the European e-Infrastructures Observatory has more than 40 indicators for monitoring the progress, usage, impact and investments of networking, high-performance computing and grids. In a few months we’ll provide the public-at-large access to an advanced set of monitoring, dissemination and collaboration tools, including regional mapping and graphs. This will help to develop better policies, and target new EU and member state investments,” said Sanchez.

Funders of e-infrastructure require this data to make the kind of computing investment needed to ensure long-term sustainability for European research.

Konstantinos Glinos, head of unit for GÉANT & e-infrastructure in Directorate General for Information Society and Media said, “powerful advanced e-infrastructures, e.g. HPC and extremely large databases, generate economic growth and improvements in the daily life of citizens.”

With an accurate cost, public-funded e-infrastructure can show their value in the face of competition from clouds and other commercial services; funders will be able to direct resources to where they are most needed.

“Based on earlier studies, the costs of the dedicated e-infrastructures seem to come out below the market price of commercial services,” said Appleton.

Glinos said, “new solutions like cloud computing and global collaboration platforms relying on public-private partnerships are being investigated to address this increasing demand for computational and data resources. It is, therefore, vital for both funding agencies and research organizations to have a clear and comprehensive understanding of the relative costs and benefits of the possible solutions, including commercial offerings, in order to compare value and deliver sustainable funding strategies.”