A scientific understanding of network designing

TL;DR

This paper establishes a scientific foundation for network design by formalizing it as a multi-objective optimization problem, analyzing the achievable design space, and evaluating different network configurations for cost-effectiveness and scalability.

Contribution

It introduces a formal framework using a Cartesian coordinate system to map and analyze the effects of various network design choices, providing new insights into optimal network configurations.

Findings

Different network types have distinct achievable design areas.

Current empirical design methods often fail to meet cost and scalability needs.

BA-like networks with efficient routing are cost-effective and scalable.

Abstract

As the Internet becomes severely overburdened with exponentially growing traffic demand, it becomes a general belief that a new generation data network is in urgent need today. However, standing at this crossroad, we find that we are in a situation that lacks a theory of network designing. This issue becomes even more serious as the recent progress of network measurement and modeling challenges the foundation of network research in the past decades. This paper tries to set up a scientific foundation for network designing by formalizing it as a multi-objective optimization process and quantifying the way different designing choices independently and collectively influence these objectives. A cartesian coordinate system is introduced to map the effect of each designing scheme to a coordinate. We investigated the achievable area of the network designing space and proved some boundary conditions. It is shown that different kind of networks display different shapes of achievable areas in the cartesian coordinate and exhibit different abilities to achieve cost-effective and scalable designing. In particular, we found that the philosophy underlying current empirical network designing and engineering fails to meet the cost-effective and evolvable requirements of network designing. We demonstrated that the efficient routing combined with effective betweenness based link bandwidth allocation scheme is a cost-effective and scalable design for BA-like scale-free networks, whereas if other designing choices cannot be determined beforehand, ER network is a markedly good candidate for cost-effective and scalable design.