论文部分内容阅读
针对传统基于BGP路由表或更新报文的路由事件识别方法由于路由更新报文噪声以及路由表采集时间间隔的限制,在路由事件识别精度和时间粒度方面存在一定局限性的问题,基于下一跳路由变化矩阵进行路由事件识别,通过BGP路由表和更新报文信息构建细粒度的路由状态变化矩阵,利用矩阵分解方法实现短时隙大规模路由事件的识别,并加以条件限制规避了影响范围较小的本地前缀事件.由于所处理的数据超过1TB,因此构建了近实时批处理的数据分析框架,并通过将此方法运用于已知的路由事件中,实验结果验证了该方法的有效性.
Traditional routing algorithm based on BGP routing table or update packet has some limitations in routing event recognition accuracy and time granularity due to the limitation of routing update packet noise and routing table collection interval. Routing change matrix routing event identification, BGP routing table and update the message to build a fine-grained routing state change matrix, the use of matrix decomposition method to achieve short-range large-scale routing event identification, and to be conditioned to circumvent the impact of more than Small local prefix event.As the data processed is more than 1TB, a near real-time batch data analysis framework is constructed, and the method is applied to known routing events. The experimental results verify the effectiveness of the method.