Abstract: A load balancing degree first algorithm on phase space for cloud computing cluster is proposed. Cloud computing cluster is of tremendous nodes and is highly coupled. The cloud computing cluster server’s parameters are casted to the phase space, and the alteration of server’s status is converted into the movements of nodes on the phase space. Load balancing degree on phase space proposed in this paper is the key parameter to evaluate cloud computing cluster’s load balancing and scheduling algorithm’s performance. So the problem of task scheduling strategy can be converted into finding the way to make the least load balancing degree on phase space. This algorithm makes the cluster’s projection on phase space gather well. Load balancing degree on phase space, board sense temperature, board sense entropy are used in the simulation experiments. These broad sense thermodynamic parameters reflect the status of the cloud computing cluster. Four kinds of experiments are designed to verify the LBFA’s dynamic and static performance. The results show that this algorithm is superior to the least load first algorithm on most performance indexes, and the load balancing degree on phase space is more stable when cluster’s scale becomes larger.