Future Telecommunications: Gedankenexperiment - Part 10 // EIT Digital

Future Telecommunications: Gedankenexperiment - Part 10

A graphic representation of Amazon Web Services (AWS) data centres and their connectivity. The future long distance network architecture may well look similar to this. The crucial difference with today's architecture is not in the wires but in how they are managed. The system is flat and completely distributed. Credit: Amazon

A crucial issue for a future network architecture will be the management of massive distributed data bases. Myriads of these data bases will be on the micro scale, let’s say in the order of a few TB each and will be owned by the owner of the devices hosting them. A smaller set of databases will be of the order of PB and EB (notice that this is “just” one thousand/one millions times larger than the ones on the user devices) and this will be owned by Cloud providers (or single companies offering data services). The balance of data between Fogs (at the edges formed by devices) and Clouds (at the edges, mostly, and formed by data centers) will be an interesting issue, particularly in a situation where there may be no centralized management (impossible? Well that’s how today’s Internet works).

The overall structure of the network will be tailored to provide efficient communications among there distributed data bases (more on this in the last post in this series).  Notice, also, that the separation between a data base as a storage facility and a processing center will fade away, both at the device level and at the Cloud level. Processing and storage can be two faces of the same coin.

The connectivity among the various data centers can be supported by a mesh of big pipes (optical highways by far) plus radio links (most of them providing connectivity to the micro data bases but also to provide connectivity in low density use areas where the deployment of a wired network is not economically viable). Storage buoys can become common place in many areas, in practice an extension of today’s Akamai mirroring infrastructure.

This kind of architecture fits well the need for data usage (and crunching). The switching in the network will not fade away but will be marginal. Already today a few of the new switches have the potential to manage a whole Countries. In Italy the current structure with some ten thousands switches could be trimmed down to some 50 new switches in a flat data centers oriented infrastructure. The deployment of these long distance fiber pipes can become embedded in the requirements for any road construction, as today you can ask for a minimum of 50mm of tarmac thickness. They are the roads for bits and bits play along with atoms also in an infrastructural way. Notice that the cost for this is marginal. Same requirements may go to the utilities deploying power lines, water pipes, sewage, urban illumination and so on. Also notice that brand new infrastructures will comply at marginal cost increase with these requirements but also existing infrastructures need to be revamped once in a while and when that happens they can accommodate bit pipes as well.

The issue of course is about the ownership of these infrastructures. Public and Private Partnership seems to be the way to go, in this particular economic framework in Western Countries but other models may well be applied. Then there is the issue of accountability (for their availability and security) but this issue is very strong in a world with limited resources. In a world of abundance the issue is basically overcome by the alternatives that are available so that a local failure does not affect the overall (connectivity) service.

What about the huge number of transactions that will be generated from IoT (as well as micro payments, on line transactions….)?

Clearly these require connectivity beyond the one provided by a wide area network. But they can piggy back on those big highways shuttling data between the Fogs and the Clouds.  

A similar point can be made for voice. Although voice already today represent a marginal part of the bits exchanged over the network(s) it remains a service that must be supported. And this service require bi-directional traffic and low latency. An architecture that is ok for broadcasting and transaction management across data centers is not the ideal one for voice. This is a specular situation of the one we have today where the voice oriented architecture is not the ideal one for data. Priority lanes shall be reserved to carry voice traffic and this should not be a major hurdles in a world of abundant connectivity. The voice architecture can be overlaid, as today we have an overlaid data architecture.

In a way the main difference of the infrastructure resulting from our gedankenexperiment is not the physical wires (or radio waves) but the way these are connected and controlled. Rather than having hierarchical structures with switching nodes as control point we have pipes connecting data centers with a completely distributed control.

Of course this begs the question: “how can an infrastructure owner charge a user?”. The answer may be shockingly simple: users are not charged! Do you get charged today when walking on a sidewalk or riding your car in a city street? You don’t. You may get charged for using specific services (special lanes, parking lots, elevators that make for an easier reach…).

The point is that in this kind of distributed architecture you also have fragmented ownership with good portion of the infrastructures deployed as public service or for different purposes (and made available for communications). Not a good scenario for today’s Telecom Operators but this is the repeating story of technology disruptions.

Author - Roberto Saracco

© 2010-2018 EIT Digital IVZW. All rights reserved. Legal notice. Privacy Policy.