Title: Graph reduction for QoS prediction of cloud-service compositions

Authors: Yanik Ngoko; Christophe Cérin; Alfredo Goldman

Addresses: Laboratoire d'Informatique de Paris Nord, Université de Paris 13, PRES Sorbonne Paris Cité, LIPN, UMR CNRS 7030, 99 avenue Jean-Baptiste Clément, 93430 Villetaneuse, France ' Laboratoire d'Informatique de Paris Nord, Université de Paris 13, PRES Sorbonne Paris Cité, LIPN, UMR CNRS 7030, 99 avenue Jean-Baptiste Clément, 93430 Villetaneuse, France ' Department of Computer Science, Institute of Mathematics and Statistic, University of Sao Paulo, Rua do Matão 1010, CEP 05508-090 São Paulo, Brazil

Abstract: This paper deals with the graph reduction approach for the QoS prediction of cloud-services' compositions. Most works with graph reduction are based on a deterministic modelling of the QoS. Though interesting, these representations are not suitable for cloud environments because they neglect the QoS variations. On clouds, we propose to use a probabilistic modelling in which the QoS of an operation is drawn from a finite set of discrete values, according to a Probability Mass Function (PMF). This paper proposes to adapt an existing graph reduction algorithm for such representations. Our first contribution is to show that there is a potential combinatorial explosion when running the reduction algorithm with the probabilistic representation. Our second contribution is to propose two heuristics in which the combinatorial explosion is controlled. The heuristics are based on two classical ideas in arithmetic: truncation and rounding. Our last contribution is an experimental comparison of the heuristics.

Keywords: service composition; cloud services; business processes; graph reduction; QoS prediction; quality of service; performance variability; modelling.

DOI: 10.1504/IJBPIM.2014.063514

International Journal of Business Process Integration and Management, 2014 Vol.7 No.2, pp.89 - 102

Published online: 21 Oct 2014 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article