中文摘要：

基于有限生灭过程建立了日历队列的数学模型,提出了一种基于马尔可夫链的动态预测算法（predictresize algorithm based on Markov,PRAM）,弥补了上述方法的不足。给出了算法的相关数学分析,并将其实现在J2EE应用服务器OnceAS中。系统实验表明,当事件到达高度密集或到达分布变化剧烈时,该算法可以解决日历队列的性能不稳定问题,使其仍保持出入队时间复杂度O（1）的特性,并且性能更优。

英文摘要：

This paper presented a new approach called PRAM,which determined the optimum operating parameter of calendar queue by predicting the future events set based on Markov chain. It implemented the PRAM prototype in the J2EE application server OnceAS. The experiment results show that PRAM offer consistent O（ 1 ） time complexity over uneven event distributions and achieve better performance than the other approaches.