Looking at they way broadband and in particular wireless networks are developing reminds me of some work I did back in the '90's on the importance of maintaining a proximity to the business. Think of a network as an eight-layer model – seven OSI layers, from the cable to the applications, plus an all important human interface which incorporates a whole host of relationships like “customer”, “business”, “consumer”. How a technology handles the eighth layer becomes all important; regardless of how removed a layer is, if you can't relate to this interface, you are simply not adding value.

This short paper by Adrian Wooster examines the impact of skipping past an understanding of customer requirements, and heading straight for network engineering heaven when developing network services, uncovering why UK telco's are falling behind in the race to better broadband along the way.

Contact us if you'd like to understand more about architecting a network for the future rather than designing for today. 

For an application developer this should be simple – they take a set of business requirements and simply translate them into an electronic incarnation. The database manager supports the application developer, so while they are a little removed, there is a good middle-man. The server manager starts to have a few problems – they are two layers removed, and their infrastructure is starting to be shared with possibly conflicting requirements. Last of all, the network manager is totally removed from the business function and is trying to manage what is, by definition, a completely shared domain with highly diverse requirements. How can the network retain some proximity to the business-end?

Plus ça change

Back in the '90's when I first looked at this, trying to link the behaviour of the network to the requirements of the applications and “human layer” was tough but achievable if you had a sizeable budget – providing you could demonstrate a solid return on investment or had a boss who liked to spend large sums on the latest toys. But deploying it required a network manager to at least make an effort to find out about active elements in the eighth layer, occasionally referred to as “customers” or even “users”, and then to work down through the layers, taking on an understanding of how the applications behaved and so forth.

No network manager ever became CIO – there's a reason. Few bothered to take even a peek up the value-chain to the people who generated the cash to pay them.

With the technology to reduce the distance between network and human layers now common-place, its clear that the understanding of how consumers work has not developed much beyond a simple graph profiling the headline traffic they generate, and the application understanding little further than a list of port-numbers to plug into the firewall.

This is as true to public broadband networks as it is of private business networks; users are a necessary evil which disrupt the smooth running of a network; new telco product sets germinate from the smallest incremental change to an existing environment and not on a commercial managers desk, and much time and effort is spent trying to control the disruptive influence of average customers.

However, taking the effort to understand the business-end pays real dividends – the network budget becomes a valued investment in the future rather than a sunk cost, and the architecture moulds itself to the business rather than expecting the users and applications to squeeze into the limitations of the network.

The Application Paradox and 3D

What I call the “application paradox” has existed since the birth of networking – this is the most fundamental base of human-aware networking. People measure the response time of off-line protocols, like email and file-transfer, in minutes and the response time of interactive applications, like voice, in milliseconds. But the inherent characteristics of any untended network will be to favour off-line “bulk transport” protocols over sensitive interactive applications. The applications people care about struggle against ones they care less about.

Acting on this, leads network managers to develop a three-dimensional view of their world. In the past technologies have been chosen for their ability to deliver great-speed, or to cover great distances – it was a simple trade-off. An understanding of the applications means technologies need to be selected from a 3D surface, where a balance is made between speed, reach, and policy.

The clever skin of a stupid network

But where does this intelligence need to be? At the edge – of course. The core network needs to be respectful but rather dumb and unquestioning – it is so far removed from the human interface that its not in a position to make meaningful decisions on anything except bit transport. However, the skin of the network, the edges exposed to the vagaries of people, needs to be sensitive and intelligent – adapting the way it accepts and transmits data, ensuring each packet gets the best start in life on its journey across the ether. The core should simply respect those decisions.

This is where the net-neutrality debate shows its true colours. By implementing the technology within the network, operators are not enhancing their ability to support the “customer” module within the human layer, because they are removing choice and dictating that specific applications, their revenue-generating applications, will behave in a prescribed manner, and everything else will suffer.

Networks have finite bandwidth, so by promoting a favoured application will necessarily impact the performance of other applications. So by promoting bulk-transport premium movies, for example, it is likely that sensitive voice traffic will become less reliable and jittery. Think of it like the bus lane on a motorway. It becomes irrelevant whether you drive a Ferrari or a Trabant, as soon as a whole lane is reserved for an infrequent, slow bus service everyone's journey will be impacted.

However, allowing applications which customers care about to have a boost over ones which they measure in minutes or hours will have no real impact on the running of the network but it may have a significant impact on the customer's experience.

The customer is king

Creating a mechanism which permits the customer to decide what's important, however, allows the person paying the piper to optimise their experience. Edge devices are necessarily in the customers domain in public networks, so the only meaningful place to employ policy is from within the customer domain; that it respects and underpins the rational power in a normal commercial relationship may be novel to the telecom's industry but its hardly revolutionary.

I can already hear the engineers mantra of “not on my network” or “but they'll abuse it”. But who, in reality, will the customer be messing up if they dump all their traffic into the platinum-with-bells-on class – pretty much no-one but themselves. Putting their P2P applications and file downloads in the same class as voice traffic will at best make phone calls suboptimal but is more likely to make them unusable, while their new movie-download will have arrived 5-seconds quicker.

The investment premium

The traditional approach which skips the bit about “requirements” and jumps straight to engineering heaven leads to short investment cycles. The UK is perhaps the archetype of the world's telecom's industry, and it is perhaps no wonder that it also has one of the shortest investment cycles, expecting to make a return on new broadband offerings in a matter of months.

Mentioning “fibre to the home” to the traditional telco's and they get hot under the collar and remind you about the billions of new investment required. For example, a figure of about £15bn is often quoted for deploying fibre in the UK; a very big figure, especially if you want to recoup that within 18-months. But if you start with a set of customer requirements and develop a structured and flexible architecture its possible to expect that investment to last many, many years.

In fact, borrowing £15bn equates to a premium of just £4 per household per month if its invested on a utility model; adding business customers and reaping the inherent cost savings from a network designed for a purpose will chip more away.

Would a consumer be prepared to pay an additional £4 per month if it took their broadband from a creaking, limiting, and controlling 2 Mbps to a liberating 1,000 Mbps, freeing them to enjoy multi-room HD TV and unlimited, unmetered teen phone lines, and whatever else they chose along the way?

Architecture is dead, long live architecture

Is seems that the dark art of the Network Architect is dead, and with it the only linkage between customers and users. Resurrecting the role will have a major impact on both the people who use computer networks, and those who invest in them.

The UK expresses some of the strongest demand-side indicators for a new broadband experience but it also has an investment community still recovering from the third-degree burns inflicted by a sector which thought it was different and doesn't need to understand their customers.

If the telcos are to break the cycle of short investment cycles and build for the future, they need to open a dialogue with the people who pay their wages – and then to architect a new generation of services.