Balanced domain division method for software defined network

Abstract

The invention discloses a software defined network oriented balanced domain division method, and aims to provide a balanced domain division method which is fast in convergence, high in efficiency and capable of reducing network control time delay and improving network performance. The method is realized through the following technical scheme: after a network topology, a node position, a pre-estimated traffic matrix and a maximum scale allowed by a single domain are given, executing an initial domain division stage based on a K-means algorithm framework, and traversing initialized distributed domain center nodes by using a width-first graph; repeatedly carrying out distance-first domain growth and domain center node updating; executing a quotient graph-based local tuning stage on the basis of an initial domain division result, converting local tuning performed by movement of domain boundary nodes in adjacent domains in the domain division result into a quotient graph, and repeatedly searching a negative weight ring or calculating a minimum cost path on the quotient graph obtained by conversion of the domain division result; and the target function is optimized and the absolute balance of the domain scale is ensured until no negative weight ring exists in the quotient graph and the domain division result reaches the absolute balance.

Description

technical field [0001] The invention belongs to the domain division field of communication network topology, and relates to division of network routing domain or management domain and clustering method of wireless ad hoc network. Background technique [0002] With the advancement of network technology and the emergence of new application scenarios, the scale of communication networks is getting larger and larger. The network scale is too large, which brings serious challenges to the scalability of the network management and control algorithms. Therefore, it is usually necessary to divide the network with a large scale into multiple smaller routing domains (or management domains) so that in network management and control. Existing domain division (or clustering) algorithms are mainly designed for traditional distributed communication networks. Due to the limitations of communication overhead, computing resources and power consumption budget, they cannot ensure the balance of...

AI Technical Summary

Problems solved by technology

Then, in the multi-objective balanced domain problem that requires the spatial distribution of nodes in the domain to be concentrated as much as possible and the traffic in the domain to be as large as possible, this condition cannot be satisfied, because the simultaneous movement of multiple domain border nodes will affect the position of the domain center node, thereby affecting the domain Calculation of the position from the border node to the center node of the domain, and the simultaneous movement of multiple domain border nodes will also affect the calculation of intra-domain traffic

In other words, directly using the increment of the objective function after moving a single domain boundary node as the weight of an edge in the quotient graph will lead to a gap between the minimum path cost on the quotient graph and the real increment of the objective function after multiple domain boundary nodes are moved. There is a deviation, and the direct consequence is a "deadlock" problem in the local optimization process, that is, node v needs to move from domain A to domain B in the current round of iteration, and node v needs to move from domain B to domain B in the next round of iteration. Domain A, the algorithm will have an infinite loop, the results of domain division will not be optimized, and the balance of domain scale cannot be guaranteed

Images