Mixed-Integer Optimization Based Green Service Selection for Large-Scale Concurrent CQA Service

Abstract

Community question answering (CQA) is widely used in Web information seeking. Web service selection (WSS) is often required to provide concurrent CQA services. WSS involves energy consumption of servers. However, existing techniques rarely consider energy consumption when dealing with concurrent WSS. To bridge this gap, we propose the mixed-integer optimization based green service selection (MIOGSS) for large-scale concurrent CQA service, which minimizes energy consumption and satisfies the common QoS of task response time. We focus on applying MIOGSS to service discovery and evaluation when dealing with concurrent request tasks. First, we propose a method to discover service candidates according to request tasks. Task de-duplication is used to reduce redundant tasks. Second, we apply a mixed integer linear programming (MILP) to evaluate service candidates by trading off their energy consumption, execution and transmission time. Third, we develop a cost model to optimize MILP, which decreases the search space of service evaluation. Consequently, a MIOGSS enabled large-scale concurrent CQA system is developed for selecting appropriate services with minimum MILP value to answer concurrent user requests. Experiments show that MIOGSS performs well in energy saving and guarantees the quality of the large-scale concurrent CQA service.