Conventional performance evaluation mechanisms focus on dedicated distributed systems. Grid computing infrastructure, on another hand, is a shared collaborative environment constructed on autonomic virtual organizations. The non-dedicated characteristic of Grid computing prevents the leverage of conventional task scheduling systems. In this study, we present the design and development of the Grid Harvest Service (GHS) performance evaluation and task scheduling system for solving large-scale applications in a shared network environment. GHS combines stochastic models and artificial intelligence learning mechanisms with task scheduling algorithms. It considers both computing and network contention and supports scheduling for single task, parallel processing, and meta-tasks. Experimental results show that GHS provides a satisfactory solution for performance prediction and task scheduling and has a real potential.