Probabilistic qualitative preference matching in long-term IaaS composition

We propose a qualitative similarity measure approach to select an optimal set of probabilistic Infrastructure-as-a-Service (IaaS) requests according to the provider's probabilistic preferences over a longterm period. The long-term qualitative preferences are represented in probabilistic temporal CP-Nets. The preferences are indexed in a k-d tree to enable the multidimensional similarity measure using tree matching approaches. A probabilistic range sampling approach is proposed to reduce the large multidimensional search space in temporal CP-Nets. A probability distribution matching approach is proposed to reduce the approximation error in the similarity measure. Experimental results prove the feasibility of the proposed approach.