Abstract: Aimed at the problem that grids are more prone to failures,and existing failure detection algorithms can not satisfy the unique requirement of grids,an efficient and scalable failure detection algorithm is then presented.According to the characteristics of grids and the small world theory,the authors established a small world based grid system model and a fault detection model;Combined unreliable fault detection method with heartbeat strategy and grey prediction model,they designed a dynamic heartbeat mechanism,and presented the efficient and scalable fault detection algorithm for grid systems further.They also analyzed the performance of the algorithm theoretically,such as how to select performance factors,as well as accuracy,completeness and scalability of the algorithm.At last,experimental result demonstrates that the algorithm is valid and effective,can be used for fault detection under grid environments.