论文部分内容阅读
提出了针对移动平台使用XMPP协议服务器端的基于分类回归树和K近邻结合的预测算法.该方法首先通过动态反馈采集服务器节点的资源信息组成时间序列,对时间序列进行预测计算.然后将服务器节点分区域管理,运用不同的调度策略.实验结果证明,与原始的加权轮询和最小连接数算法相比,该预测算法在连接响应时间上减少了25%,在建立连接的平均速率上提升了近1.3倍,动态的调度策略使得服务器集群有更大的吞吐量,对于移动平台有更好的适应性.
This paper proposes a prediction algorithm based on the combination of classification and regression tree and K-Nearest Neighbor for XMPP protocol on mobile platform.Firstly, the dynamic feedback is used to collect the resource information of server nodes to compose the time series and predict the time series.And then, Regional management and different scheduling strategies.Experimental results show that the proposed algorithm reduces the connection response time by 25% compared with the original weighted-round-robin algorithm and the minimum connection number algorithm, and improves the average speed of establishing connection 1.3 times, dynamic scheduling strategy makes the server cluster has greater throughput, better adaptability for mobile platforms.