一种基于两级存储结构的网络流量测量算法

Flow Traffic Measurement Algorithm Based on Two-layer Memory Hierarchy

为了准确而完备地测量高速骨干网中各条流的流量,需要容量大且速度快的存储器来保存所有流的状态信息,因而代价极高。该文提出了一种基于两级存储结构的网络流量测量算法。两级存储结构由容量小但速度快的一级存储器和容量大但速度慢的二级存储器构成。考虑到网络流量分布的Quasi-Zipf法则,测量算法尽量将大流量流的状态信息保存在一级存储器中,将小流量流的状态信息保存在二级存储器中,较好地解决了存储器容量和速度之间的矛盾。仿真结果表明,与抽样测量相比,该算法具有较小、较平均的测量误差。

Explicit measurement of per-flow traffic is difficult in backbone networks because it needs large high-speed memories. The main contribution of this paper is a new flow traffic measurement algorithm based on two-layer memory hierarchy. Such memory hierarchy is constructed by small high-speed memory on the first layer and large low-speed memory on the second layer. Illumined by Quasi-Zipf＇s law of network flow size, the measurement algorithm inclines to save the state information of flows with heavy traffic in layer-one memory and that of flows with light traffic in layer-two memory. The two-layer memory hierarchy makes a better tradeoff between space and speed compared with large high-speed memories. It shows experimentally that the algorithm proposed in this paper has a smaller and fairer estimation error.