Within service-oriented Grids, resources may be advertised and traded as services and participants need to have economic incentive. Grid can be seen as a multi-agent system, agent are rational and state their QoS requirements in the form of a utility function which maximize is his goal. In this paper, we propose utility-based computing model for grid. We consider the problem of task scheduling subject to time and budget constrains as joint utility function, and scheduling object is to maximize total attained utility. This schedulingproblem can be shown to be NP-hard, thus we present a joint utility function scheduling algorithm (JWS) for this problem. The performance of model is analyzed by SPNP sofmare package. The numerical results show that JUFS can provide user more utility.