Networks as “Optimizers” just emerging from micro-scale rules ?Thursday, June 14th, 2012 by Antonio Manzalini
In previous posts, we’ve elaborated how a network can be seen as an emergent property of a complex ecosystem. Networks emergence is generally describes as macro-scale properties resulting from micro-scale rules. In the prior-art analysis of network layering, we often find that layers networks can be defined as optimizers maximizing specific composed/aggregated utility functions. If we integrate these two perspectives, in a broader sense, we can define networks as emergent property of sets of optimizers gaming to maximize specific utility functions.
I can see this for natural ecosystems. Imagine a swarm of fishes or even better a nest of ants: the “communication” networks between ants is an emerging property (producing self-organization) to optimize both the “life” of any single ant (which may have its own local utility function) and the organization overall community (having an aggregated utility function for the nest). Actually, I’ve never heard about any risk of “Tragedy of Commons” in ants’ nets: evolution has selected those autonomic behaviors to avoid “breakdowns”, by keeping a delicate equilibrium of local vs global utility functions. A lesson from Nature on self-organization to be learnt.
Can we approach future networks, as ecosystems of resources, in the same way ? Which are the utility functions we can design for controlling a communication network ? One can image functions about sources rates, useful information, delay, energy consumption…etc.
In this direction, F. Kelly, has shown that TCP/IP protocol is a perfect example of optimizer: its objective is to maximize the sum of source utilities (as functions of rates) with constraints on resources. And actually, each variant of congestion control protocol can be seen as a distributed algorithm maximizing a particular utility function. The exact shape of the utility function can be reverse engineered from the given protocol. Similarly, Border Gateway Protocols (BGPs) can be seen as a solution to the Stable Path Problem, and contention-based Medium Access Control (MAC) protocols as a game-theoretic selfish utility maximization. Other utility functions could be User satisfaction (e.g. User-generated pricing following end-to-end principle), resource allocation efficiency or different network economics fairness.
Then, modeling networks as emergent property of a sets of optimizers, means considering management-control based on interacting controllers maximizing a combination (e.g. weighted sum) or an aggregation (e.g. in multiplicative form) of several utility functions. Or, alternatively, we may say that management-control should look for the network Pareto optimality, or it should play an uncooperative dynamic game. In any case, this would imply looking at future networks management-control with a different perspective, through the glasses of a deep vertical and horizontal network decomposition.
The emerging paradigm of Software Defined Network (SDN) is about having a fully decoupled network control plane, so it can be seen from this broader perspective, at least at the edge of current infrastructures. In a SDN, control intelligence is (logically) centralized in software-based controllers. Said controllers provide visibility and control over the network, they can ensure that access control, routing, traffic engineering, QoS, security, and other policies are enforced consistently across the network infrastructures. Governing the interactions of these controllers would allow managing and optimizing a SDN according to certain policies, or utility functions.
In other words, approaches like SDN seems paving the way to look at future networks in a different way, as ecosystems of resources, where top down governance (e.g. playing the role of evolution ?) of sets of controllers could meet emergent properties from local bottom-up autonomic behaviors (e.g. via local utilities).
But let’s go even beyond this (partly) engineered approach: can we build only on bottom-up emergent properties (just based on micro-scale rules) to get self-stabilizing future networks ecosystems, indeed like in Nature ? A great challenge towards 0-Capex, 0-Opex networks.