The centralized radio access cellular network infrastructure based on centralized Super Base Station(CSBS) is a promising solution to reduce the high construction cost and energy consumption of conventional cellular networks. With CSBS, the computing resource for communication protocol processing could be managed flexibly according the protocol load to improve the resource efficiency. Since the protocol load changes frequently and may exceed the capacity of processors, load balancing is needed. However, existing load balancing mechanisms used in data centers cannot satisfy the real-time requirement of the communication protocol processing. Therefore, a new computing resource adjustment scheme is proposed for communication protocol processing in the CSBS architecture. First of all, the main principles of protocol processing resource adjustment is concluded, followed by the analysis on the processing resource outage probability that the computing resource becomes inadequate for protocol processing as load changes. Following the adjustment principles, the proposed scheme is designed to reduce the processing resource outage probability based onthe optimized connected graph which is constructed by the approximate Kruskal algorithm. Simulation re-sults show that compared with the conventional load balancing mechanisms, the proposed scheme can reduce the occurrence number of inadequate processing resource and the additional resource consumption of adjustment greatly. The centralized radio access cellular network infrastructure based on centralized super base station (CSBS) is a promising solution to reduce the high construction cost and energy consumption of conventional cellular networks. With CSBS, the computing resource for communication protocol processing could be managed flexibly according to the Since the load transfer mechanisms used in data centers can not satisfy the real-time requirement of the communication protocol processing. First of all, the main principles of protocol processing resource adjustment is concluded, followed by the analysis on the processing resource outage probability that the computing resource becomes inadequate for protocol processing the load changes. The proposed scheme is designed to reduce the processing resource outage probability based on the optimized connected graph which is constructed by the approximate Kruskal algorithm. Simulation re-sults show that compared with the conventional load balancing mechanisms, the proposed scheme can reduce the occurrence number of inadequate processing resource and the additional resource consumption of adjusted greatly.