Quality of Service （QoS）-based service selection is the key to large-scale service-oriented Internet of Things （lOT）, due to the increasing emergence of massive services with various QoS. Current methods either ha...Quality of Service （QoS）-based service selection is the key to large-scale service-oriented Internet of Things （lOT）, due to the increasing emergence of massive services with various QoS. Current methods either have low selection accuracy or are highly time-consuming （e.g., exponential time complexity）, neither of which are desirable in large-scale lOT applications. We investigate a QoS-based service selection method to solve this problem. The main challenges are that we need to not only improve the selection accuracy but also decrease the time complexity to make them suitable for large-scale lOT applications. We address these challenges with the following three basic ideas. First, we present a lightweight description method to describe the QoS, dramatically decreasing the time complexity of service selection. Further more, based on this QoS description, we decompose the complex problem of QoS-based service selection into a simple and basic sub-problem. Finally, based on this problem decomposition, we present a QoS-based service matching algorithm, which greatly improves selection accuracy by considering the whole meaning of the predicates. The traces-driven simulations show that our method can increase the matching precision by 69% and the recall rate by 20% in comparison with current methods. Moreover, theoretical analysis illustrates that our method has polynomial time complexity, i.e., O（m^2 × n）, where m and n denote the number of predicates and services, respectively.展开更多
基金supported by the National Natural Science Foundation of China (Nos. 61272487, 61232018, and 61402009)
文摘Quality of Service （QoS）-based service selection is the key to large-scale service-oriented Internet of Things （lOT）, due to the increasing emergence of massive services with various QoS. Current methods either have low selection accuracy or are highly time-consuming （e.g., exponential time complexity）, neither of which are desirable in large-scale lOT applications. We investigate a QoS-based service selection method to solve this problem. The main challenges are that we need to not only improve the selection accuracy but also decrease the time complexity to make them suitable for large-scale lOT applications. We address these challenges with the following three basic ideas. First, we present a lightweight description method to describe the QoS, dramatically decreasing the time complexity of service selection. Further more, based on this QoS description, we decompose the complex problem of QoS-based service selection into a simple and basic sub-problem. Finally, based on this problem decomposition, we present a QoS-based service matching algorithm, which greatly improves selection accuracy by considering the whole meaning of the predicates. The traces-driven simulations show that our method can increase the matching precision by 69% and the recall rate by 20% in comparison with current methods. Moreover, theoretical analysis illustrates that our method has polynomial time complexity, i.e., O（m^2 × n）, where m and n denote the number of predicates and services, respectively.
