IP offload: Why not pipe it ALL onto the Internet?

Getting ready to nit-pick the idea? OK, that's your right. But my plea is to just give it a bit of a mental hearing first and then, and only then, reject it out of hand with reasons for doing so in the comments box below.

Here's the full proposal and - before you write anything dismissive - yes, I'm sure it or something like it has been considered many times before in many a bar after many a conference (it does seem pretty obvious). If so, sorry. But here is the idea again with extra, current reasons for considering it now.

It's this: if femtocells can work technically and in business model terms by offloading data 'on the 'user side' of the 3G air interface into and across (both offload and breakout) the Internet, why can't the same approach work on the other side? Why not treat the RAN (radio access network) as a big femto - a macro femto, if you like - and just hand off the whole data backhaul and core network part of the piece to broadband Internet access providers?

I've been mulling the question for a while and recently I bounced it off "offload" expert Manish Singh, VP of Product Line Management at Continuous Computing.

Manish is a gentleman and is always prepared to humour me on the telephone for far longer than my ideas and observations deserve.

He agreed that the concept of employing local IP access and local IP breakout had been gaining ground fast recently because of the smartphone (and soon to be iPad) data crunch which is gobbling up available mobile broadband capacity at an alarming rate. Home-based WiFi tied to broadband access can be engineered to pipe data across the Internet to the mobile core network, thus helping to de-congest both the licenced radio access network and the backhaul network set up to serve it. This is a big trend at the moment, Manish explained.

A similar benefit is accrued by femtocells which essentially use the operator's licensed spectrum to do the same thing as WiFi. Like WiFi, femtos sit in the home and use the Internet to connect to the core operator network and, also like WiFi, are increasingly designed to discriminate between traffic which needs to be sucked into the operator core - either because it's part of the network's signalling plane or requires special QoS - and traffic which is simply heading to the open Web and can be 'broken out' to travel there under its own steam.

In addition, Manish told me that offload activities are happening around the edge of the network as operators increasingly seek to use breakout techniques to relieve congestion there. The idea is to take the top off the data demand and to release the data to the Internet 'earlier' (or receive it 'later' depending which way around you're thinking about it). This, he said, not only saved costs and reduced congestion, it could also improve performance for the breakout Web apps which no longer have to multi-hop through the core mobile network before being released to the Internet.

OK, so lets start drawing some strands together.

One of the key dynamics here is the proportion of traffic making a needless journey through the telco core network just to be spat out to the Internet on the other side. There are variable stats on this. I've heard it can be as high as 95 per cent of the 'bits'. Manish says it might be 60-70 per cent of the traffic. Either way it's over half and, as video and other bandwidth hungry apps increase, can only grow faster.

So that's a huge proportion of traffic which doesn't need to go to the operator core but in many cases does - and it includes much of the new 'smartphone boom' traffic generated by video, Facebook, Youtube and so on.

So one way of tackling the bandwidth crunch is to drive new revenues to pay for scaling up the mobile network RAN and core. Higher charges and charging per bit, for instance; perhaps working on regulators to allow upstream charging of content providers.

Alternatively, new architectures might be dreamt up to radically reduce the costs of deployment and mesh with the new reality of mobile broadband. That reality is that 3 and 3.5G (LTE) will not turn on a money tap for operators by enabling all sorts of paid-for services. Mobile broadband will not foster a 'Mobile Internet'. Rather it is already enabling mobile access to the fixed Internet for services with a mobile twist, and there is absolutely no sign that this will change.

In fact the idea that IMS (IP Multimedia Subsystem) and RCS (Rich Communications Services) are going to usher in a new world of mobile telco-delivered is now dead - so why are operators still factoring in the idea that they need to keep scaling up the core (and scaling up all the expensive boxes that reside within it) to cope with this development?

Surely the optimum network for this wave of traffic would be one which simply offloaded ALL the data at the RAN onto a broadband access circuit? No backhaul, no core network.

 

Web traffic (content and signalling) could be handed straight over to the Internet and signal traffic (for user validation, security, data counting and so on) could be tunnelled across the Internet to a server controlled by the network operator.

Yes, cell hand-off might be a problem. But then again, for data and video applications with all the buffering already in place to allow for variable web delivery, users would be unlikely to notice slow hand-off.

For voice, of course, this architecture could be a problem. But then many operators envisage their LTE networks initially being deployed for data-only anyway, with voice being served by Circuit Switch (CS) fallback, so no loss there.

Manish is sceptical but, as I say, he's a gentleman. I ask him whether he's ever been approached by a carrier to look into developing this sort of architecture. I'm particularly thinking about low-cost operators. These guys have been brilliant at deploying GSM service in developing countries at far lower cost than has apparently been possible in developed markets. Does it not make sense that when it comes to deploying 3G and LTE data services, such an architecture could at least be considered?

Manish hasn't heard of it, but he agrees such an approach could be a direction of travel undertaken by an operator at some point in the future.

What do you think?