The doctoral student from the Universitat Politècnica de Valencia (UPV) and Universitat Jaume I in Castelló Antonio J. Peña caught the interest of the participants in the SuperComputing 2011 exhibition held in Seattle in mid-November with the presentation of a new technology, rCUDA, which allows remote access to graphics accelerators in a set of high performance computers (HPC cluster).
The application was developed by the doctoral student in the research currently being undertaken for his doctoral thesis at the Universitat Jaume I (UJI), directed by the lecturers Federico Silla, from the Parallel Architectures Group in Valencia, and Rafael Mayo from, the High Performance Computing and Architectures Group in Castelló, who also participated in the demonstration.
The proposed system consists of a structure of graphic cards to support networked computing in complex operations to hundreds or thousands of nodes that make up a high-performance computer, thus getting 100% efficiency. This system saves energy by using fewer graphics processors and reduces investment in equipment and maintenance costs.
The demonstration at the supercomputing fair has been made possible thanks to the invitation of the company Mellanox Technologies, the most important company in the implementation of InfiniBand technology, a protocol open to any company and provider of services for servers and storage. The company considers that this technology is highly innovative and has provided material for research free of charge. Besides, the company has invited researchers to participate in several conferences.
The Parallel Architecture Group in Valencia and the High Performance Computing and Architectures Group in Castelló are working together on this technology for over two years. The working group rCUDA is made up by Jose Duato, Javier Nadal and Carlos Reaño from the UPV, and Enrique Quintana and Adrián Castelló from the UJI, apart from the doctoral student and the directors of his thesis.
The computing world is working to achieve in 2020 the first computer with exaflow capacity that will make 10 raised to the power of 18 operations per second and will solve calculations of issues related to climate change or genomic sequence to which current technology can not response. Between the technologies used to develop this new computer there is the GPU computing (using the graphics processing unit to perform scientific computing).