Currently, the combined push-pull model is usually applied in the resource monitoring under cloud computing environment. However, the quantity of data gathered by the producers is much less than those received by the monitor so that a lot of resource is wasted. In order to maintain the data consistency of real time system, an improved adaptive algorithm based on combined push-pull model is proposed. The adaptive strategy is added on the producers that aims to predict when refreshes data came from the monitor using the history data. Experimental result shows that the algorithm can decrease the resource consumption while exchanging data effectively.