Supercomputing for small companies made simple
In the third report from the PRACEdays15 meeting in Dublin at the end of May, Tom Wilkie discusses how supercomputing can be made simpler for smaller companies.
The most fragile part of the HPC ecosystem is the network of small firms, spin-out companies and open source software service companies – and it is a major task for the community to find ways to nurture and support them, the PRACEdays15 meeting in Dublin was told at the end of last month.
The problem is that such companies do not have critical mass and are often geographically dispersed – so there could be a role for the Partnership for Advanced Computing in Europe (Prace) as an ‘incubator’ for such companies and for the ‘early adopters’ of HPC among small to medium enterprises (SMEs), Lee Margetts, lecturer in computational mechanics at the University of Manchester, told his audience.
A special session of the meeting was devoted to demonstrating how these smaller companies in the HPC ecosystem could encourage and support the wider use of high-performance computing in industry and also for some unexpected and interesting applications in the public sector.
In a hugely dynamic and enthusiastic presentation, Stefano Cozzini, CEO of eXact Lab in northern Italy, offered some ideas and practical examples of how to achieve these objectives. His company exists to ease access to high performance computing for SMEs and for the public sector as well.
It started as an HPC consultancy, but after three years in business, Cozzini and his co-founders realised that consultancy ‘does not scale’ and that they needed a scalable solution to provide their services. This led to the eXact computing environment, targeting SMEs that have limited or no HPC infrastructure and for whom, although they have simulations already in place, their HPC needs still have to be properly identified. These could range from CFD to rendering for media companies, he said.
One fruit of their labours is XeRis – a cloud platform for advanced analysis of seismic hazards. The potential user can get an assessment of the safety of buildings during earthquakes through a web-based interface. Among the users has been a Swiss nuclear power plant. But it has also found application in the public sector, with the Government of Trieste in Italy using the service to assess the seismic safety of schools in the region. (As the destruction of Assisi due to two devastating earthquakes that shook Umbria in September 1997 testifies, earthquakes are a real risk in Italy.)
A second project is OpenViewShip, an industrial/academic project to conduct hydrodynamics simulations for the naval industrial sector in the Veneziano region. A critical element to this project was remote visualisation so that large data files did not need to be transported. In the end, according to Cozzini, the actual computational performance was not improved, but the productivity was increased because remote visualisation had eliminated data transport.
But despite the energy and enthusiasm of Cozzini’s presentation, he admitted that there were challenges. The first was to find enough people who were trained in high-performance computing and the second was establishing networks of small and medium sized companies who could provide the expertise needed.
In an effort to address the first of these problems, SISSA and ICTP are running a Masters programme in high-performance computing. In early 2012, a group of Slovenian and Italian companies came together to form the High Performance and Cloud Computing Cross-Border Competence Consortium (HPC5) to provide businesses, researchers, and universities with advanced HPC and cloud computing services so as to encourage, facilitate and enhance research and innovation in technologically intensive vertical niches and stimulate growth in the region. HPC5 is dedicated to the creation of the most innovative, creative and powerful centre for HPC and cloud computing services in the Slovenia-Italy cross-border region.
Manuel Arenaz, from the Spanish company Appentra Solutions, reminded his audience that ‘programming supercomputers is hard’. His company, a spin-off from the University of Corunna, has therefore developed a software tool to add parallelisation to sequential codes and to do this almost automatically.
The technique employs pragmas in OpenMP to do the translation from sequential to parallel. A pragma or directive is a language construct that specifies how a compiler should process its input. Directives are not part of the language proper but instead specify the compiler’s behaviour.
The company has developed a specific product which it is calling ‘Parallware’ to automatically find course-grain parallelism in the sequential source code, without intervention by the programmer.
Speaking by video link as he was unable to attend the session in person, Arenaz placed his company’s work in the context of a long history of global efforts, both in the academic and industrial sectors, to build parallelising compilers that convert sequential scientific programs into their parallel equivalents. However, even such examples as PGI and GCC compilers, he maintained, suffer from problems as a result of their development having started in the 1970s and 1980s which creates limitations when applied to real codes for today’s applications. The traditional approach is based on classical compiler theory for data dependence analysis, and idiom recognition and is sensitive to variations in the syntax and quality of different implementations of a given algorithm.
Arenaz sees Parallware as a way of addressing the HPC ‘talent gap’ because it should allow engineers and scientists to focus on their science and engineering and decouple that task from the details of the underlying parallel hardware.
HPC on demand
A different approach to making access to HPC easier for small and medium sized companies has been taken by Constelcom, which is addressing not so much the problem of developing software that will run efficiently on multiply parallel machines but ways of getting round the high capital cost of high performance computers. Nicholas Tonello, the company’s director, outlined his proposal for what is in effect a self-managed HPC cloud for simulation and collaboration. The first production version will go live at the UK’s Hartree centre this autumn, he said. The constellation cloud platform for multiphysics, HPC and collaboration will allow users to access the HPC resources quickly and simply – effectively on-demand – rather than users having to estimate how much time they would need and apply for it in advance. In this way, the platform would put the user rather than the machine at the centre of the process, he claimed.
This is the third in a series of reports from PraceDays15, held in Dublin last week. Robert's Roe's report on the role of HPC in the host country, Ireland, can be found here. European supercomputing policy was set out at PraceDays15, prompting the question whether a European company will build Europe's first Exascale computer. The reasons why parallel programs need new maths are explained by Tom Wilkie in a further report from the conference. In the final report, Robert Roe compares and contrasts the support given for HPC in Japan and in Europe.