Task Scheduling is a critical design issue of distributed computing. The emerging Grid computing infrastructure consists of heterogeneous resource in widely distributed autonomous domains, which makes task scheduling even more challenging. In this paper, a general resource scheduling algorithm for computational grid, OD (On-Demand) is present. According to the computational grid job?s personal resource requirements, the different importance of these requirements and the system load of the computational grid nodes, OD chooses an appropriate node for computational grid job from the candidates finally. The experiments show the performance improvement for applications and achieve better load balance by this algorithm.