Intradomain routing optimization based on evolutionary computation
Vítor Pereira · Portuguese National Funding Agency for Science, Research and Technology (RCAAP Project by FCT) · 2019
The growing number of Internet-connected devices and the escalation of new Internet services, such as cloud services and video streaming, are some examples of factors that increase the volume and mutability of traffic in communication networks. The need to channel increasingly large volumes of traffic in network infrastructures with limited capacity, highlights the importance of Traffic Engineering mechanisms that aim to deliver an efficient use of network resources. Routing decisions play a key role as they de ne how traffic is distributed on the available paths and hence how networks resources are explored. Traditional routing protocols, such as Open Shortest Path First and Intermediate System to Intermediate System, have constraints which prevent an optimal network resources utilization. Some of these constraints are, for example, their inability to perform uneven splitting of traffic across multiple paths and their lack of a centralized control. These constraints impose additional difficulties when changes in network operating conditions need to be considered, such as significant variations intra c necessities and link failures. When facing such changes, routing configurations need to adapt to the new conditions and ensure that the network continues to operate efficiently. Communication technologies are constantly evolving. Recently, alternative routing solutions have emerged that enable new Traffic Engineering approaches. Network management problems and, in particular, the optimization of network resources utilization can be addressed using such alternative solutions. Software-De ned Networking and Segment Routing paradigms provide greater exibility in the selection of routing paths, and although they can overcome many of the constraints of traditional routing protocols, it is necessary to find configurations, both scalable and manageable in real contexts, that optimize the distribution of available resources. In this context, this research work intends to respond to the stated problems by proposing efficient mechanisms for optimizing the use of resources in networks configured with traditional Link State routing protocols, as well as in networks that implement the latest paradigms of Software De ned Networking and Segment Routing. In addition to enabling efficient resource utilization, the proposed optimization mechanisms are responsive to relevant changes in the network environment that may result from variations in traffic requirements or topology changes such as link failures. The nature of the problems, which besides not being solvable in polynomial time include more than one optimization objective in their formulation, are addressed using algorithms fostered in the field of Evolutionary Computation. Distinct traffic requirements and different network states frequently require clashing configurations. Evolutionary Algorithms possess several characteristics that are desirable to solve problems with multiple conflicting goals and make them preferable to classical optimization methods. They provide a set of compromise solutions to problems for which there is no single optimal solution. The research work winded up in an autonomous optimization framework that integrates all the proposed Traffic Engineering methods, which is made publicly available to be freely used by researchers and network administrators.