MWC 2012 Network Tech: Edge Power, Super WiFi, Van Jacobson
It's been Mobile World Congress again. To tide you over until we've finished the traditional period of fasting, sleeping, and using no technology invented more recently than the sundial, we've put together a mini-series of blog posts about major trends we noticed. This one covers what's happening in networks and infrastructure: Edge Power; Super WiFi; and Telco 2.0's Top Network Tech moment - a Living Legend on QoS vs. QoE.
(NB For further Telco 2.0 research on future broadband strategies see 'The value of "Smart Pipes" to mobile network operators', and our broadband research stream.)
Last year's network buzzword was "het-net", and there was plenty of that about, but this year's was "small cells". The original femtocell concept foresaw a device with a similar form factor and deployment pattern to a SOHO wireless access point, but this ran into quite a few problems, notably that not many people actually saw a need to have one in their home, that the relationship with other network operators was never really worked out, and that the smartphones are always more than happy to move over to the wireless LAN. WLAN equipment being cheaper by an order of magnitude and more common in the home and office by several orders of magnitude, it got used. Also, deploying significant numbers of devices that wouldn't be radio-planned was a problem.
Small cells are a difference of emphasis rather than kind - typically, they are used to provide public rather than homezone coverage, they are available in more sizes (anything less than a full scale macro-cell, really), and they usually pack more features.
But it's worth looking through the radio issues. Ubiquisys' new range of small cells, for example, have moved over to Intel x86 hardware in the interests of cost and also of general-purpose programmability. Using PC-like hardware means that it should be much easier to develop applications to run on them. They also typically include a large solid-state disk, like the ones in Apple MacBook Airs, which provides mass storage with very fast input-output.
(Source: Ubiquisys on Flickr)
The obvious use for this is to integrate content-delivery networking as deep into the radio network as it's possible to get - in fact, you could just make out command-line output in Keith Day's slides logging cache-hits and misses for YouTube videos. The former Femto Forum, now the Small Cells Forum, is working on a standard API for these devices in order to facilitate building services like operator CDNing.
In the same field, but more radical, is Quortus, a British startup that implements mobile core network functions in software in order to port them to devices other than the giant switches they usually run on. That could mean bog standard IT gear, which in any case some networks already run on, or it could mean femtocell or PC-scale devices pushed out to the network edge.
This has a huge range of potential applications. For example, it makes it possible to build a network with a distributed architecture, spreading the core functions and data around and therefore avoiding the need for expensive specialised gear, reducing single points of failure, and eliminating sources of network latency. One reference application is a rural deployment, using big high power radios for coverage, but getting rid of the high capacity RNCs that go with them and instead using a Quortus node to control the radios. Being an MSC in its own right, speaking SIP and MAP to the rest of the network, it can keep providing local calls if the backhaul goes down.
Alternatively, such a device could be used as part of an enterprise network, providing in-building coverage and also acting as a fully mobile PBX for advanced voice features. We note that PABX capability is on the Small Cells Forum roadmap, as well. When you look at what Vodafone has achieved with One Net, it should be clear that there are big possibilities here.
Another element of this is, of course, WLAN. Our Managed Offload use case sounded radical when we published, but as one of the GSMA panellists said, "operators have discovered that WiFi is something they can trust" and now they can't get enough of it - a quick-deploying source of high speed data connectivity with zero spectrum costs and no core-thrashing signalling load. Operator WLAN is surging up onto the productivity plateau, and as a result, vendors like Ruckus Wireless and (inevitably) Cisco will be taking more of the telcos' capex dollar in future.
Many of the same points about edge power apply to WLAN equipment as well. A lot of small cell products have a WLAN radio as an option, and the Wireless Broadband Alliance (WBA) at least talks a good game about additional services and applications via the Hotspot 2.0 initiative.
However, the biggest barrier to moving more data traffic off the cellular network is the same one that has always dogged WLAN. If the authentication is secure, it's not usable, and if it's usable it's not secure - and to be honest, most landing pages aren't very usable. As a result, the WBA is trying hard to get more SIM-based automatic authentication (EAP) deployed, and some big names are expected to announce progress quite soon. Although EAP only helps devices with a SIM, this may not be as big a problem as all that - after all, smartphones and tablets are the main drivers of traffic growth.
Tellingly, even the conference WiFi worked this time.
A Living Legend On QoS vs. QoE
Telco 2.0's Top Network Tech moment came rather discreetly in an anonymous, little advertised seminar room deep in Hall 5 on Wednesday afternoon, while the world's attention was focused on Microsoft's lavish Windows 8 preview and trying to get invited to the Google party.
As a graduate student back in the late 1980s, Van Jacobson solved the congestion collapse that crippled the nascent Internet, by inventing TCP/IP's congestion management features. If it hadn't been for that, the US academic and military research communities might have given up on IP and used something else (a lot of networking technology was tried and abandoned in the 80s and 90s), and the world would be very different. (Actually it might be even more different had he left a toilet out of a classic 1999 paper.)
Later, he helped to start Cisco Systems, invented traceroute and the Multicast Backbone, before returning to research at Xerox PARC.
Someone to listen to, then. But we didn't know that he studied FedEx's network of trucks moving around the world looking for insights into networking in general.
"I asked FedEx how they managed to make money when they were stuck in the same traffic jams I was. The answer was that there are traffic jams, but a lot of the time, there aren't, and FedEx operates 24 hours a day and never loses an opportunity to move a packet."
Van also thinks a lot of our problems are down to the way we've ended up using the Internet. It's a medium that is designed around one-to-one interaction, via the famous end to end principle, and it's often running on top of access networks designed for one-to-one phone calls...but it's mostly used for either one-to-many broadcasting (like the top 10 on YouTube) or many-to-one aggregation (like thousands of coders throwing commits into Github, or hundreds of thousands of photographers pouring images into Flickr).
His proposed solution is to embed content and applications into the Internet, bringing it much closer to the user. Rather than a cloud solution, though, he argues that this should be much more distributed, with applications and caches in every network domain. (The potential significance of those SSDs and Intel Atom processors in small cells ought to be clear at this point.)
And, fascinatingly, he thinks that there is a lot to learn from logistics companies like FedEx in general. Faced with widely diverse, shared, and unreliable transport links of different speeds, they manage to provide reliable delivery and differentiated service, without moaning about "over-the-top transport providers (i.e. cars) hogging our roadwidth".
Van also pointed out that there's not much money in owning a road. Perhaps there's a vision here - of a shared or public macro-infrastructure, with differentiated providers of digital logistics operating on it, and a wide range of different access technologies at the edge.