数据广播调度自适应信道划分与分配方法

随着移动网络的不断发展，移动终端设备的计算能力与日俱增，越来越多的用户倾向于通过移动网络获取信息资源，这使得实时按需数据广播面临新的挑战：（1）数据内容和规模的多样化；（2）用户请求的实时性与需求多样性使得热点数据增加，直接导致广播数据总量的剧增；（3）用户对服务质量和水平的要求越来越高.当前的研究成果主要集中在固定信道模型和算法上，一定程度上忽略了当前数据广播调度环境的变化.固定信道存在如下问题：（1）局限于特定的网络，缺乏通用性；（2）信道大小、个数不能随着网络环境的变化而自动调整，降低了广播效率.基于以上考虑，对实时按需数据广播调度的自适应信道划分和分配进行研究，提出一种自适应信道划分与分配方法OCSM （optimized channel split method），其根据数据请求特征的不同，实时自适应地调整信道个数和大小，从而提高系统敏感性、鲁棒性以及广播效率.该方法包括：（1）广播数据均衡聚类算法WASC （weight average and size clusteralgorithm），其挖掘数据特征，为信道划分提供依据；（2）数据项广播优先级评定算法R×W/SL，其实时评定数据项调度优先级；（3）信道划分与分配算法CSA （channel split algorithm）.实验包括两个方面：（1）确定不同数据项大小和请求截止期分布下的信道划分策略，并分析聚类算法中聚类距离K在不同情况下的最佳取值以及最佳信道划分；（2）验证自适应信道划分与分配策略的有效性，并通过对比实验验证在不同情况下OCSM的有效性.实验结果表明：OCSM优于其他调度算法，并具有较强的自适应.

Channel Split and Allocation Method for Data Broadcast Scheduling

With the rapid development of mobile networks and a great increase in the computing ability of mobile devices, a huge number of people tend to obtain information through mobile networks, which poses some new challenges for real-time on-demand data broadcasting:(1) The data types and sizes are diverse; (2) The real-time characteristics and demand diversity of the user requests greatly increase the volume of hot-spot data (the most access data) and the volume of broadcast data; (3) The users'' demands for high service quality become stronger. Current research has been focusing on the fixed-channel models and algorithms and ignoring the changes of real-time data broadcast environments. The problems of fixed-channel models are as follows:(1) They are limited to specific network with fixed channel-models which lack generality; (2) The size and number of channels cannot be adjusted with the changing of broadcast environments automatically. This paper studies the possibility of an automatic channel split and allocation method that can adapt to the environment, and proposes an optimized channel split method (OCSM), which can adjust the bandwidth and number of broadcast channels to the different characteristics of real-time requests. The method includes the following algorithms:(1) A weight average and size cluster algorithm (WASC) for data characteristics mining; (2) A weight evaluating algorithm (R×W/SL) for evaluating the priority of data item; (3) A channel split algorithm (CSA) for channel split. The experiments undertaken in this study include two aspects:(1) Determining the different strategies under different data size distributions and deadline distributions; (2) Verifying the validity of OCSM by validating the effectiveness in different situations through a series of experiments. The results reveal that significantly better performance can be obtained by using the OCSM rather than other state-of-the-art scheduling algorithms.