Data optimised computing for heterogeneous big data computing applications

The rise of big science techniques is reshaping the provisioning of computing resources and scientific software in large science facilities. As facilities are gearing up for data intensive computing infrastructure, a wave of facility-based big science computing platforms is emerging. This paper presents a new computing paradigm towards designing HPC data analysis platform, named Data Optimised Computing (DOC). The DOC paradigm leverages the characteristics of science data to optimize HPC resource utilization and to improve users´ ability to harness a variety of scientific analysis software frameworks. We present a preliminary architectural design of a software platform that implements this approach and also discuss the future directions of this work.