一种面向大规模网络仿真的自适应拓扑划分机制

An Adaptive Topology Partition Mechanism for Large Scale Network Emulation

随着网络空间靶场等技术的发展,大规模网络仿真的需求越来越迫切。受限于主机的处理能力,大规模网络仿真需要在多台主机上部署,而如何合理对网络拓扑进行划分就成为影响仿真效率的关键因素。为此,提出了基于物理仿真主机性能约束的自适应拓扑划分模型,并在典型拓扑划分工具METIS的基础上实现了相应的划分算法。拓扑划分算法包括仿真规模预估和拓扑划分调整两大步骤,前者根据网络拓扑和物理主机性能约束估算需要划分的子块数目,后者根据物理主机性能约束进行拓扑的划分和调整。实验结果表明所提出的拓扑划分机制能够有效针对大规模网络的仿真需求进行拓扑划分,并保证划分结果满足物理主机性能约束和子块间负载均衡等目标。

With the development of network space range technology, the requirement on large scale network emulation becomes more and more serious. Constrained by the host capacity, the emulation of a large scale network should be deployed on many hosts simultaneously, and thus how to partition a network topology effectively will affect the emulation efficiency dramatically. In this paper, it proposes a physical emulation host capacity constrained adaptive topology partition model, and implement the corresponding partition algorithm based on a topology partition tool METIS. The proposed algorithm includes two steps: namely emulation size estimation and topology partition and adjustment. The former estimates the number of partition parts according to the network topology and the physical host capacity constraint, while the latter performs the topology partition and adjustment based on the estimation result. The experiment results show that the proposed mechanism can partition the topology of a large scale network effectively, and can guarantee the aims of satisfying the physical host capacity constraint as well as load balancing.