The invention provides a multi-resource scheduling method oriented to resource cost optimization in cloud computing. The method comprises the following steps of step A, obtaining an application submitted by a user; and step B, determining optimal resource allocation according to the application submitted by the user. An SmartYARN algorithm provided by the invention is capable of favorably solving the problem of optimization on multi-resource allocation in cloud computing, i.e., on the premise of ensuring the performance constraint set by the user, the expenses for task resources of the user are minimized by virtue of allocating reasonable cloud computing resources. The method is applicable to batch processing tasks with large data volume, long running time and resource expense decrease expectation; a large number of experiments indicate that the method not only is capable of achieving the actual optimization accuracy rate of 98%, but also is relatively low in performance expenses, and is generally capable of finding the optimal resource allocation applicable to the task only through 15-step search.