Probabilistic top-K dominating services composition with uncertain QoS

Abstract

Traditional service selection schemes require users to define a utility function by assigning weights to each quality-of-service (QoS) metric. To relieve users from the professional knowledge, skyline techniques have been studied recently by several researchers. However, the size of skyline services is sometimes not easy controlled due to intrinsic attributes of services. Additionally, we observe that most QoS metrics may fluctuate during run-time. Considering such uncertainty and dynamics, in this paper, we propose to obtain probabilistic top-k dominating services with uncertain QoS. Different from previous works, our approach employs the probabilistic characteristic of service instances and calculates the dominating abilities of services so as to achieve an accurate selection. Experimental results have shown the feasibility and effectiveness of our approach.