CERIT Scientific Cloud (CERIT-SC)
Centre CERIT-SC (CERIT Scientific Cloud) is a national centre operating a computing and data storage experimental infrastructure for research and development in the area of flexible e-infrastructures and large in-silico experiments, performed in close collaboration with other scientific disciplines. Its top-level mission can be phrased as “Speeding up the time from ideas to publications (and products) in all research disciplines”. The centre is built on three pillars: hardware resources of sufficient scale to ensure competitiveness, excellent research in specific areas of computer science (i.e. know-how to use the hardware resources efficiently), and long-term collaboration with the user communities (being true research partners of the users, not just providers of “precanned” technical solutions). This project aims at strengthening the first and the second pillar; the third one is currently being funded by other means.
Specifically, significant part of project budget is investment to hardware, pushing the centre’s equipment to the leading edge of available technology. This is complemented with the in-house research programme, with the goal to increase the efficient usage of the current and procured infrastructure.
The research programme will consist of two integrated subprogrammes aimed at big data analysis and high performance computing. Their approach to efficient use of infrastructure is complementary. While the Big Data subprogramme approaches it more top-down, concentrating on data processing methods and software architectures at a higher level, the High Performance Computing subprogramme concentrates on optimization at a lower level, from bottom-up. The Big Data research subprogramme will focus on the research in efficient big data analysis, with special focus on the careful selection of the best-fit data analysis technique to each specific problem setup. The High Performance Computing research subprogramme will focus on software parallelization, acceleration and optimization, including automated frameworks.
Total number of publications: 20