Skip to main content
Kent Academic Repository

Complex System Modelling for Communication Networks

Marshall, Ian W. (2002) Complex System Modelling for Communication Networks. In: Complexity Modelling Workshop. . (The full text of this publication is not currently available from this repository. You may be able to access a copy if URLs are provided) (KAR id:13709)

The full text of this publication is not currently available from this repository. You may be able to access a copy if URLs are provided.

Abstract

It is often said that one of the most complex machines built by mankind is the global telecommunications network. Certainly if one’s definition of complexity accounts for number of components (billions), number of nodes (~100k) or degree of connectivity (2-2000) the communications network is extremely complex. It should therefore not be surprising that it exhibits a variety of “symptoms” of complexity including power law connectivity (Barabasi), small world traverses, Fractal Traffic (Bestavros), emergent behaviours (AT&T Brownout), indeterminate fault conditions, etc.

Historically, modelling of communications networks has been based on assumptions of predictable behaviour, including Poisson traffic that aggregates to a smooth continuum in the core network, linear scaling rules, and independent faults. Clearly these assumptions are over simplifications, but whilst the network was mainly offering telephony, and was only growing slowly, they gave results that were useful. A large body of tools and expertise therefore built up using these simple assumptions. However, the emergence of the Internet has pushed network operators into an era of rapid (exponential) growth of capacity (that exposes the limitations of simple scaling rules), orders of magnitude more burstiness (that can no longer be treated as representing exceptional transients), and much more complex interdependence relationships between services (more than just telephony). Network operators are therefore starting to explore novel approaches to system modelling based on the ideas developed in the complex systems community in the last 20 years.

For example it is now routine to include a fractal traffic generator in many models to assess sensitivity to burstiness, and multi-agent or cellular automata models are common in advanced telecommunications research. There are also a few groups (e.g. Dorigo, Marshall) exploring models of self-organizing networks, protocols and management systems. Acceptance of the results by operational engineers has proved to be a major barrier however. Reasons include:

1. The techniques are largely statistical in nature and do not offer rigorous proofs of stability or absolute bounds on behaviour,

2. The mode of thought required is novel and many engineers feel threatened by the implied need to retrain, or simply prefer to stay with the tools they know,

3. Engineer’s and manager’s intuitions about complex network systems are often incorrect, and at variance with the model results.

In contrast, we have found that intuitions regarding natural complex systems are often highly sophisticated. We are therefore making extensive use of natural analogies in our current research. However care must be taken to avoid confusion – many biological analogies have associations with disease or pests, and arouse unintended emotional reactions that distract the audience and reduce the chances of acceptance. A more neutral source of analogy is in geophysical systems, such as rivers, metamorphic reaction fronts, oceanic currents etc. As a result our current research borrows heavily from geophysical analogies. For example we are developing energy efficient routing algorithms for mobile ad hoc sensor networks based on tidal currents, algorithms for autonomous placement of network services based on hydrothermal alteration at diffusion reaction fronts, and algorithms for prioritizing traffic based on fractional crystallization. We believe that our approach will be of interest to other scientists working in these and related areas, and that there may be mutual benefits to be obtained from future collaboration.

Item Type: Conference or workshop item (Paper)
Subjects: Q Science > QA Mathematics (inc Computing science) > QA 76 Software, computer programming,
Divisions: Divisions > Division of Computing, Engineering and Mathematical Sciences > School of Computing
Depositing User: Mark Wheadon
Date Deposited: 24 Nov 2008 17:59 UTC
Last Modified: 16 Nov 2021 09:51 UTC
Resource URI: https://kar.kent.ac.uk/id/eprint/13709 (The current URI for this page, for reference purposes)

University of Kent Author Information

Marshall, Ian W..

Creator's ORCID:
CReDIT Contributor Roles:
  • Depositors only (login required):

Total unique views for this document in KAR since July 2020. For more details click on the image.