The invention discloses a dynamic resource scheduling method for large-scale computing clusters. The method comprises the steps of 1) presetting maximum and minimum quantities of computing resources available for each resource user; 2) collecting a total current available computing resource quantity wn, a quantity of computing resources occupied by each resource user and a total current resource user quantity; and 3) if free idle computing resources exist currently, allocating the free computing resources to the resource user with an operation demand, and if the idle computing resources do not exist and the resource user with the operation demand exists, extracting a plurality of the computing resources from the computing resources used by each resource user currently or gradually reducing the available computing resource quantity of the current computing resource user and allocating the extracted computing resources to the new resource user with the operation demand. According to the method, the overall resource utilization rate of the clusters is increased; and especially when a large amount of single-core serial operations exist in the clusters, the effect is very remarkable.