A tutorial on using genetic algorithms for the design of network topology

Abstract

SUMMARYThe design of network topology is an important part of network design, since network topology is directly associated with network operational behavior, capacity, reliability, and cost. This paper is a tutorial paper concerned with illustrating how the optimization capabilities of genetic algorithms can be used to design suitable network topologies considering basic topology problems. Simple genetic algorithms have been developed for the topology problem of mesh networks, considering single node and single link failure tolerance. The algorithms are based on criteria of two important measures: minimizing the length of communication links; and minimizing traffic flow through these links for given traffic loads. The first measure contributes to minimizing the cost of cabling, while the second measure contributes to minimizing the cost of link capacity. The work provides a useful approach and tools to network students and professionals concerned with the topology design of backbone networks. The developed software is made available on the Internet.

Fiber optical network design problems: A case for Turkey

“…However, due to curse of dimensionality, problems may be difficult to solve [32]. Also, other techniques are used as simulated annealing [33,34], tabu search [35], evolutionary algorithms [36,37,38], ant colony optimization [39] and heuristics [40,32]. For earlier studies about the backbone network design, the reader can refer to [41,32,42,43,44,45,46].…”

Strategies on 3-connected graphs optimization search for future ICT infrastructure

“…The topologies used are the following: DR, CR (20,5), CR(20, 9), N 2R(10, 3), and D3 for the topology found using the newly introduced approach. CR(20, 3) is equivalent to DR and CR(20, 7) is equivalent to CR(20, 5).…”

“…The combination of links problem is itself NP hard [5], and in addition, the 3-connected constraint (or minimum cut calculation) implies another problem to be solved in polynomial time for each of the iterations in the search process [6]. The main goal is to present and evaluate novel techniques and procedures to design degree 3, 3-connected graphs without the adjacency matrix as an input.…”

Network Planning Using GA For Regular Topologies

“…Evolutionary Algorithms (EAs) and in particular Genetic Algorithm (GA) have successfully been applied to solve constrained problems with multi-objectives, such as transportation problems [7], production process planning problems [8] and network topology design problems [1]- [3], [9] and [10]. EAs/GAs were investigated for several kinds of encoding methods [8] where most of them can not effectively encode/decode without getting some infeasible solutions that require some repair before being considered.…”

“…Once the number of nodes increases (n > 20), optimization of the network characteristics (total length, maximum link length, availability, budget, etc) for, even, a known topology can be challenging for any human or dedicated algorithms except perhaps for generalized heuristics algorithms such as EAs/GAs where the objective is optimization of the characteristics of these topologies even thought that there is no guarantee that the best solution can be found, nevertheless the guarantee is that after many runs better solutions can be found [1] and [2]. …”