Résumé: Cloud computing has become one of the most popular frameworks for big data processing and analysis. To have better performance, the data chunks and their corresponding process nodes must be near each other, which is known as data locality. Moreover, to meet the deadline constraints, the available processing power next to the data locations is crucial. Considering the time and resource constraints, designing an appropriate job scheduler is essential for the effective management of memory and processing capacity. These concerns coalesce into the problem of joint data placement and job scheduling with data locality and deadline constraints. In this paper, two kinds of job sets are determined, one with equal deadlines and the other one with various deadlines. The problem of joint data placement and job scheduling with a single deadline is formulated as the problem of bin packing with splittable items and cardinality constraints, which is strongly NP-hard. Moreover, this paper proposes the polynomial-time Meddal algorithm to find a feasible solution with a few processing nodes. Experiment results have shown that the proposed method is effective, with the number of required nodes reduced by when compared to the algorithms Cred and First Fit.

Langue: Anglais