内容分发网络中基于请求的对象一致性算法

Request-Based Object Consistency Algorithm in Content Distribution Networks

在内容分发网络中,为了保证对象新鲜的同时又不给网络和Web服务器增加过多的负荷,提出了一种全新的基于请求的对象一致性算法(ROCA).它要求复制服务器以每个独立的对象为单位,按照一定的刷新间隔从Web服务器上下载对象的最新版本.为了确定对象的请求频度,定义了对象兴趣度,并以此为依据确定对象的刷新策略.对于兴趣度高的对象采用滑动时间窗的方式,根据上一个时间窗内的对象请求和修改情况,对时间窗尺寸进行动态修正.基于轨迹驱动的仿真实验表明,ROCA算法能够根据对象特征快速地调整刷新间隔取值,与传统算法相比,它具有较高的请求新鲜度和较低的带宽浪费率.

In order to ensure the object＇s novelty while increasing no more load to Web servers and networks, a novel request-based object consistency algorithm （ROCA） for content distribution networks is proposed, in which regarding every independent object as a unit the duplication of servers is required and the newest version of the object downloaded from Web servers is updated according to certain time interval. To determine the frequency of the object＇s request the interest-measure of the object is defined and the update policy of the object is decided based on it. For the object with higher interest-measure the time slipwindow method is utilized and the size of time window is dynamically corrected in terms of the request and corrected instances of the object within the previous time window. Trace-driven experiments show that ROCA algorithm can regulate the update interval quickly according to the characteristics of the object.Compared with traditional methods it has higher novelty of object and lower bandwidth wastage.