Network segmentation optimization algorithm applied to ns3 parallel simulation

Abstract

The invention provides a network partitioning optimization algorithm applied to NS3 parallel simulation. The algorithm comprises the following steps: S1, acquiring a traffic matrix by analyzing Netflow data; reading a routing table in a database to generate a routing matrix file; S2, generating link traffic and node traffic, generating link traffic through the routing matrix file and the traffic matrix, and combining the link traffic with a link connection state file to generate the node traffic; S3, generating a CSR format file of a corresponding Metis partitioning algorithm, generating the CSR format file of the corresponding Metis partitioning algorithm by standardizing the link weight and the node weight. By using the optimization algorithm provided by the invention, the acquired Netflow traffic data is processed, the traffic data of the node and the link is computed by using a traffic matrix generation method so as to acquire the weight of the node and the link; the original network topology is partitioned according to the weight of the node and the side by using the Metis algorithm so as to balance the simulation task load of each partition, thereby improving the time efficiency of the whole NS3 parallel simulation system.

Description

technical field [0001] The invention belongs to the technical field of computer simulation, in particular to a network segmentation optimization algorithm applied to NS3 parallel simulation. Background technique [0002] Computer simulation refers to the calculation and simulation of the actual or virtual system behavior model within a certain period of time. Since the 1960s and 1970s, with the continuous development of science and technology, the models established by human research in many fields have become larger and larger, such as: complex networks, weather forecasts, new drug development, intelligent machines, aircraft test flights, oil exploration, Real-time economic planning, nuclear physics, and plasma physics, etc., require an astonishing increase in the amount of calculations required. Ordinary serial simulations take too long to complete and cannot meet people's needs more and more. Therefore, parallel simulations have become an important means of computing and ...

AI Technical Summary

Problems solved by technology

However, since the Netflow that collects SNMP data and generates tasks belongs to two different modules, it cannot be synchronized in a strict sense, that is, to ensure that the generated weights and simulation tasks are in the same time period

This will lead to some errors in the division results generated by the division algorithm, resulting in unbalanced division results and excessive communication between divisions, which will seriously affect the final parallel simulation efficiency and fail to achieve the expected parallel simulation optimization. Effect

Images