Fast-FFA: a fast online scheduling approach for big data stream computing with future features-aware

摘要

Awareness of future features is more important than that of historical features for online scheduling in a big data stream computing environment. In this paper, a fast future feature-aware online scheduling approach fast-FFA is put forward, exhibiting the following contributions; 1) Modelling the online resource scheduling from viewpoints of user and data centre, considering multi-dimensional features of online data stream and quantitating preferences and utilities of each dimension. 2) Obtaining future features from historical features of multidimensional data stream with a hybrid particle swarm optimisation, back propagation (PSO-BP) algorithm and optimising online scheduling with an immune clonal algorithm. 3) Evaluating fast-FFA and balancing both fast future feature awareness and acceptable accuracy objectives. Experimental results demonstrate that the proposed fast-FFA approach has high potential as the approach provides significant system efficiency enhancements in online big data environments.