| 摘要: |
| 针对高速骨干网流量分析与流量管理对大规模高速统计计数器的需求,提出了一种简单、高效的主动式计数器结构——DALCA,其新颖之处在于对计数器向量进行分层,除第1 层外,其他各层子计数器向量采用多级哈希表的形式进行组织,获得了良好的空间效率和时间效率.与现有的主动式计数器结构BRICK 相比, DALCA 的空间效率与其接近,但是时间效率明显优于BRICK.采用真实的骨干网流量数据,通过仿真实验对DALCA 的性能进行了评估,结果表明:执行查询操作时,DALCA的访存带宽需求约为BRICK 的1/10;执行更新操作时,DALCA的访存带宽需求约为BRICK 的1/6. |
| 关键词: 骨干网 流量分析 流测量 统计计数器 多级哈希表 |
| DOI：10.3724/SP.J.1001.2013.04365 |
| 分类号: |
| 基金项目:江苏省自然科学基金(BK2010103) |
|
| Statistics Counter Architecture for Backbone Network Traffic Analysis and Management |
|
ZHANG Jin, HUANG Qing-Shan, ZHAO Wen-Dong, PENG Lai-Xian
|
|
Institute of Communication Engineering, PLA University of Science and Technology, Nanjing 210007, China
|
| Abstract: |
| A simple and efficient active statistics counter architecture named DALCA (direct addressing layered counter array) is presented for high speed network traffic analysis and management. The novelty of DALCA is that the counter vector is split up into multiple layers, and all layers except the first one are organized as multi-level hash tables. This makes DALCA efficient in both from the space and time perspective. DALCA has similar space efficiency and significantly higher time efficiency compared with BRICK;the state-of- the-art active statistics counter architecture. The performance of DALCA is evaluated using real world backbone network traffic traces. Simulation results show that the memory bandwidth demand of DALCA is about 1/10 and 1/6 that of BRICK during query and update operations respectively. |
| Key words: backbone network traffic analysis flow measurement statistics counter multi-level Hash table |
