How Would Ants Implement an Intelligent Route Control System?

Abstract

Multihoming, the connection of a stub network through multiple Internet Service Providers (ISPs) to the Internet, has broadly been employed by enterprise networks as a sort of redundancy technique to augment the availability and reliability of their Internet access. Recently, with the emergence of Intelligent Route Control (IRC) products, IRC-capable multihomed networks dynamically select which ISPs’ link to use for different destinations in their traffic in a smart way to bypass congested or long paths as well as Internet outages. This dynamic traffic switch between upstream ISPs is mostly driven by regular measurement of performance metrics such as delay, loss ratio, and available bandwidth of existing upstream paths. However, since IRC systems are commercial products, details of their technical implementation are not available yet. Having the incentive to delve into these systems deeply, in this paper, we employ traditional ant colony optimization (ACO) paradigm to study IRC systems in that domain. Specifically, we are interested in two major questions. Firstly, how much effectively does an ant based IRC system switch between upstream links in comparison to a commercial IRC system? Secondly, what are the realistic underlying performance metrics by which ants pick the path to a food source (destination network) in a multihomed colony? Through extensive simulations under different traffic load and link reliability scenarios, we observe that ants perform well in switching between available egress links. Moreover, delay of paths is not the only criterion by which ants select the path; instead, through their intuitive ACO paradigm, they tend to choose the path with a better performance in terms of both delay and loss ratio.