Over the last few years, service-oriented Grid that enable a new way of service provision based on utility computing models, which users consume based on their QoS (Quality of Service) requirements. In such services grids, resource management and scheduling based on user's QoS constraints are addressed seldom, especially in the aspect of users' multiple QoS attributes preferences considered synthetically. This paper proposes a novel scheduling algorithm based on integrated utility of multiple QoS attributes. It describes multiple QoS attributes as utility functions and drives the dynamic resource scheduling by taking these utility functions as objective functions of scheduling optimization. Considering tasks related features, this paper also brings forward a task partition execution strategy to improve scheduling efficiency. Experiments show that the scheduling algorithm can reflect user's utility and preference in general. It also provides services division and performance guarantees, which presents the on-demand service well and makes scheduling more flexible.