Solving the multicommodity flow problem using an evolutionary routing algorithm in a computer network environment

PLOS ONE RESEARCH ARTICLE Solving the multicommodity flow problem using an evolutionary routing algorithm in a computer network environment Noel Farrugia ID*, Johann A. Briffa ID, Victor Buttigieg Department of Communications and Computer Engineering, University of Malta, Msida, Malta * a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 OPEN ACCESS Citation: Farrugia N, Briffa JA, Buttigieg V (2023) Solving the multicommodity flow problem using an evolutionary routing algorithm in a computer network environment. PLoS ONE 18(4): e0278317. https://doi.org/10.1371/journal.pone.0278317 Editor: Ping He, Jinan University, China, HONG KONG Received: March 25, 2022 Accepted: November 15, 2022 Abstract The continued increase in Internet traffic requires that routing algorithms make the best use of all available network resources. Most of the current deployed networks are not doing so due to their use of single path routing algorithms. In this work we propose the use of a multipath capable routing algorithm using Evolutionary Algorithms (EAs) that take into account all the traffic going over the network and the link capacities by leveraging the information available at the Software Defined Network (SDN) controller. The designed routing algorithm uses Per-Packet multipath routing to make the best use of the network’s resources. PerPacket multipath is known to have adverse affects when used with TCP, so we propose modifications to the Multipath TCP (MPTCP) protocol to overcome this. Network simulations are performed on a real world network model with 41 nodes and 60 bidirectional links. Results for the EA routing solution with the modified MPTCP protocol show a 29% increase in the total network Goodput, and a more than 50% average reduction in a flow’s end-to-end delay, when compared to OSPF and standard TCP under the same network topology and flow request conditions. Published: April 19, 2023 Copyright: © 2023 Farrugia et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Data Availability Statement: The relevant data are available in the GitHub and Zenodo repositories (https://github.com/um-dsrg/era-routing); (https:// github.com/um-dsrg/era-lemon); (https://github. com/um-dsrg/era-ns3); (https://zenodo.org/record/ 6783634). Funding: This research has been carried out using computational facilities procured through the European Regional Development Fund, Project ERDF-076 ‘Refurbishing the Signal Processing Laboratory within the Department of CCE’, University of Malta. NF was granted the Introduction One of the major drawbacks of computer networks using a distributed architecture is their low efficiency caused by the lack of routing solutions aware of the entire network status. In a distributed architecture network, every routing device contains its own independent control plane; each routing device takes independent routing decisions based on information local to the device. A distributed network can improve the routing decisions taken by individual components by constantly sharing a snapshot of the current global network status. However, due to the impracticality of such a solution, distributed network architectures resort to either single path routing algorithms or very simple multipath solutions such as Equal Cost Multipath Routing (ECMP) [1]. In ECMP, flows are distributed over paths with equal cost where a hash of the packet header, not the current network state determines the path taken [2]. Improving the network resource usage efficiency requires the use of routing algorithms with access to an accurate and up-to-date snapshot of the global network status. A centralized network PLOS ONE | https://doi.org/10.1371/journal.pone.0278317 April 19, 2023 1 / 28 PLOS ONE ENDEAVOUR Scholarships Scheme (Group B). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Competing interests: The authors have declared that no competing interests exist. Solving the multicommodity flow problem using an evolutionary routing algorithm architecture, such as that provided by a Software Defined Network (SDN) [3], provides such information by moving the control plane to a central location. While SDN and control plane centralization is often considered a recent invention, the centralization of the network control plane is anything but; [4] gives an interesting insight on the origins of SDN. As an example, the improved efficiency enabled by SDN has been vital in the deployment of Google’s B4 international network that would have otherwise been unfeasible [5]. In this work, we aim to increase the network’s efficiency by focusing on route optimization for an already deployed network. An alternative would be to design a network topology from scratch that lends itself to the use of multipath and reduced power consumption to operate such a network. Topology design is beyond the scope of this paper; the interested reader is referred to [6] for a survey on specific network topology design and the routing algorithms designed specifically for them. The problem of routing network traffic intelligently based on the current network status to meet a number of user defined objectives falls under the class of optimization problems known as the Multi-Commodity Flow Problems (MCFPs). A MCFP is an optimization problem whereby a given set of flows are routed over a given network topology to optimize for a number of objectives while adhering to a set of constraints. Most commonly, solutions to the MCFP are found using Linear Programming (LP). LP solvers are guaranteed to find the optimal solution based on the constraints and conditions set; however, LP solvers are unable to optimize for multiple objectives. Such solvers are also restricted to the use of linear objectives and constraints as otherwise finding a solution becomes an NP-hard problem. As we will see later, the linearity condition, is too restrictive to accurately model some of the network’s behaviour dealt with here. For example, constraining the MCFP such that a flow may only traverse over a single path, is a non-linear constraint and converts the problem to an NP-hard one when solved using LP [7]. When designing routing algorithms, one rarely needs to optimize for a single objective, so that, the MCFP becomes a multi-objective optimization problem and a multi-objective solver is required. Two very common, often conflicting objectives, when dealing with network routing, which are the ones used in this work, are throughput maximization and delay minimization. With a high enough demand, an optimal solution to the MCFP (without the single path constraint), will make use of multipath routing. Multipath routing is key to make the best usage of the available network resources [8, 9] (...truncated)