Web Service Composition with Uncertain QoS: An IQCP Model

Abstract

Quality of Service (QoS) is commonly employed to represent nonfunctional web service (WS) characteristics for the purpose of optimizing WS composition. As a departure from most of the extant research on QoS aggregations, where QoS is typically represented deterministically, we hypothesize the QoS to a WS as a random variable that follows a normal distribution. A serial of formulas are proposed to calculate the expectation and variance of the QoS of a composite service; this yields four QoS criteria suited to workflow described by a directed acrylic graph (DAG). The Web service composition problem with uncertain QoS is then modeled as an integer quadratically constrained program (IQCP). Finally, a series of experimental results obtained in CPLEX and Java illustrate that our model has favorable robustness and can estimate composite service QoS rapidly and accurately.