Data-centric Computing on Distributed Resources

Distributed computing has always been a challenge due to the NP-completeness of finding optimal underlying management routines. The advent of big data increases the dimensionality of the problem whereby data partitionability, processing complexity and locality play a crucial role in the effectiveness of distributed systems. The flexibility and control brought forward by virtualization means that for the first time we control the whole stack from application down to the network layer but, to a certain extent, the best way to exploit this level of programmability still eludes us.Our research tackles this problem from both the data and the infrastructure fronts. We investigate the evolving dynamic infrastructure whereby we research distributed computing on inter-clouds and web browsers. Dynamism in the infrastructure leads to more adaptable middleware; we investigate prediction and fuzzy based data processing scaling techniques for workflows and dataflows. The increasing complexity in data processing is a challenge; we address this complexity by introducing an automata-based modeling and coordination system.The role of semantics will play an essential role in the future of data processing. For this reason we investigate how semantics can be used to inter-link globally distributed data processors. This layer forms the final layer in our research which started from the dynamic infrastructure layer.