One of the main objectives of BSC-CNS is to proactively transfer technology to industry, both as an objective in itself in terms of dissemination of scientific output, and also with the intention to generate industrial returns.
Technology Transfer can occur in many ways, including the publication of research results in academic as well as industry journals, licensing of proprietary technology (which is usually protected via patents or copyright) to private industry, spin-off of technology and know-how into a start-up company, joint collaborations with industry in applied research and development of new techniques and products, training of scientists and technicians who then go on to work for private industry, and direct training of workers via short courses, workshops, etc.
It is generally very difficult to patent advances in computer coding and chip design and this limits the opportunities to directly license or spin-off new supercomputing technologies. Although the BSC-CNS does do this when possible, the main focus in technology transfer is in the training of people and joint R&D.
In 2010 the Operations Department expanded its training courses for both RES users and operators of RES nodes, and continued its work to offer improved levels of user support to assist users in overcoming technical issues.
Joint collaborations with private industry are undertaken at both the system level, via direct collaborations with industry leaders such as IBM and Microsoft, and at the application level, where a significant research effort is dedicated to providing supercomputing based models for solving engineering problems in industrial sectors such as aerospace, transport, energy, medicine, geology, etc.
MareIncognito is a bilateral IBM-BSC research project to define the hardware and software components for the new generations of Petascale supercomputers, under the umbrella of a memorandum of understanding, led by Mateo Valero and Jesús Labarta. It encompasses research in several fields related to supercomputing, including: application porting and novel numerical methods, programming models and their runtime implementation, performance analysis and prediction tools, resource management layers, interconnection networks, and memory and processor architecture for novel multicore-based architectures. During 2010, the Project has been redefined in three major research areas: architecture, performance and workflows. A set of SOWs for specific work within each research area have been defined, and plan to start during 2011.
The Kaleidoscope Project, led by José María Cela, is a "dream team" partnership of top geophysicists, computer scientists and organisations from around the world. It has been initiated by Repsol YPF, a Spanish integrated oil company with large assets in the US Gulf of Mexico, 3DGeo, a leading Houston-based imaging company formed by Stanford University professor and seismic imaging pioneer Biondo Biondi, and the BSC-CNS. The Kaleidoscope Project has privileged access through the BSC-CNS to Cell/BE based systems and technology because the BSC-CNS is one of the few research centres in the world developing libraries and codes for such processors.
The Kaleidoscope Project aims to produce more reliable and faster (by several orders of magnitude) software tools to analyse geo-seismic data and visualise below the thick layers of salt present in the Gulf of Mexico. This will significantly reduce exploration risks and make accessible oil reserves that otherwise would be invisible to the industry.
BSC-Microsoft Research Centre
The BSC-Microsoft Research Centre, led by Osman Unsal and Adrián Cristal, was established in April 2008 to focus on the way in which microprocessors and software for the mobile and desktop market segments will be designed and interact over the next 10 years and beyond. The presence of many- and multi-core processor computing architectures has make it possible to deliver enormous computational power on a single chip, with profound implications for the way software is developed. Optimising the design and interaction of hardware and software architectures to take advantage of the new computing power will require tight integration across the industry.
Computer architecture experts at BSC-CNS are teamed up with computer scientists at Microsoft Research Cambridge (in the United Kingdom) and Redmond (in the United States) to look for innovative solutions to the challenges and opportunities that massively parallel processing entails. The vision of the centre is of a top-down computer architecture in which software requirements drive the hardware innovation forward rather than letting the hardware design condition software development. Fundamental and applied research is being conducted in the following main topics: hardware/software transactional memory, runtime systems to support dataflow paradigms in novel architectures, OS/architectural support for managed programming languages and vector architectures for low-power media devices.
Nvidia-BSC CUDA Research Center
The BSC-CNS has been named by NVIDIA as a 2010 CUDA Research Center, the first one in Spain. The CUDA Research Center Program recognizes and fosters collaboration with research groups at universities and research institutes that are expanding the frontier of massively parallel computing. The Center, lead by Prof. Nacho Navarro, recognizes BSC’s broad-based research success in leveraging the NVIDIA CUDA technology and GPU computing, both in major research initiatives such as Global Memory for Accelerators (GMAC) and in projects dealing with the optimisation of applications, such as Reverse Time Migration (RTM). The CUDA Research Center also recognises BSC-CNS’s efforts in CUDA education, highlighted by means of the 2010 Summer School, “Programming and tUning Massively Parallel Systems (PUMPS),” offered in July on the Barcelona university campus. The event was co-sponsored by the University of Illinois, the HiPEAC NOE and NVIDIA, with distinguished faculty members Dr. David B. Kirk of NVIDIA and Prof. Wen-mei Hwu of the University of Illinois. PUMPS attracted more than 100 attendees from throughout the EU and beyond, from beginners to advanced faculty. The Center will promote the training in programming languages like CUDA, OpenCL, OpenMP, and StarSs, and the efficient optimisation of runtime environments and numerical methods.
Other Industry and Institutional Collaborations
Goal: to provide computational resources, including 1 Million CPU-hours in MareNostrum, to ITER, the EU´s hydrogen fusion reactor project.
Goal: to study the impact of natural gas-powered vehicles on the air quality in Barcelona and Madrid.
CISCO Systems Inc.
Goal: to model multicore multithreaded architectures in typical network applications and estimate the performance of different architectures.
SGS TECNOS S.L.
Goal: the goal of this collaboration was to perform an impact assessment on the air quality from combined cycle power stations.
Goal: to perform air quality modelling.
Goal: to perform a simulation for the Air Quality Plan in Catalonia.
AEMET (The State Meteorological Agency)
Goal: to implement, disseminate and validate the operational prediction of the North African
dust transport in the Iberian Peninsula as well as to perform modelling, detection, follow-up
and characterisation studies of atmospheric material.
Goal: to develop and implement an operational high-resolution air quality forecasting system for Spain, providing end-users with an air quality forecasting and assessment service for Spain and Europe with higher detail for some hot spot areas.
Government of the Canary Islands
Goal: to develop an information system about the air quality forecast and surveillance of the Canary Islands.
Goal: to provide the Andalusia Government with an operational air quality forecasting and assessment service, which will allow the simulations of photochemical and particulate matter pollution with high-spatial and temporal resolution for Andalusia: 1 km2 and 1 hour.