I’ve just spent a couple of days at the first offload-specific conference I’ve come across, organised by IIR. It’s been useful, giving me some good new contacts and allowing me to reconnect with some existing clients and friends. This blog post is just a summary of some of my take-outs and reflections – some people may already have seen some tweets I posted with the #mobileoffload tag.
One thing that seems to be coming through quite strongly is that WiFi offload is currently taking the high ground in comparison to femtocells, especially in the residential marketplace. Conversely, there seems to be growing momentum for outdoor “small cells” of various types compared to WiFi hotzones.
Neither of these are absolutes, but these fit into a general narrative that:
- Outdoor coverage and capacity is “classical” mobile network territory in terms of both personnel, planning and operator processes. Consequently, the idea of small cells fits with the notion of “more of the same, but smaller and cheaper and easy to site and manage”.
- Indoor data offload tends to be driven by consumers’ familiarity with – and often preference for – WiFi. The heaviest mobile data users are almost certainly a subset of those that are happy with the setup and operation of WiFi in their homes. Operators' ability to deal with WiFi's vagaries through device-side software or inbuilt standards is also improving.
The conference focused heavily on this last point - what I'm calling the “telco-isation” of WiFi. There are various standards and specifications being worked on by the WiFi Alliance, Wireless Broadband Alliance, 3GPP, GSMA and others. There’s an alphabet-soup of acronyms here – 802.1x, 802.11u, Hotspot 2.0, ANDSF, I-WLAN, WISPr and plenty of others. There was lots of talk of EAP-SIM authentication and so-called "seamless" mobility. There are various approaches to dealing with the operators' own and partnered WiFi accesses, especially around extensions of the mobile roaming mechanisms.
Some of this is very important and is being done intelligently and effectively. The idea of improved "network discovery" so that you can tell more about a WiFi access than it's SSID name makes a lot of sense. It is also important in some cases for operators to be able to "steer" users to particular APs or SSIDs, and collect information and maybe enforce certain policies. In some cases, SIM-based authentication can make sense as well - an area where my opinion has shifted a bit recently.
But however – and this is a big however – I think there are some serious issues. In my view, the industry is in danger of making the same mistakes it made with UMA about 5 years ago. The giveaway is in this cliched word "seamless". I spent a lot of timing criticising this aspect of UMA, and I can see myself having the same conversations all over again. Seamlessness is not the utopian ideal, just as it wasn't in 2006.
In a nutshell - sometimes, and for some use-cases - automated and seamless (ie zero user-touch) connection to WiFi is absolutely desirable, ideally with session continuity and all that other fine stuff. But, critically there are also various use cases where seams are important, and need to be made visible to the user and/or applications running on the device. The tricky part is designing the end-to-end system, and especially the user interface on the connection manager, to cope with both sets of scenarios.
Seams might be "messy", but they are appropriate for certain contexts. To reiterate the analogy I made in my presentation, we don't all go around wearing Lycra catsuits. Our clothes still have seams for good reasons, and the same is true of networks. Ultimately a seam is a border, at which things change - speed, latency, security, cost, ownership, policy, power consumption and many other parameters. The idea that the border should always be crossed with the user kept unawares risks a whole host of problems.
There are various angles here:
- The user will often wish to connect to WiFi networks that are not "approved" or linked to the operator's network or WISP partnerships. Most obviously, the user will want to use home broadband WiFi, private enterprise networks (often behind a firewall and with the corporate network's own security and authentication) and free public WiFi where it is available. The operator-driven WiFi software must not get in the way of this type of scenario - and neither should it try to tunnel back via the operator core in these cases.
- The user may have access to multiple WiFi networks in a given location. The operator-preferred one may not be the best - perhaps because it costs more, perhaps because it is slower, perhaps because policies are enforced that the user would rather were not. Auto-connecting anyway may be an undesirable outcome.
- The same WiFi network may be available locally on better terms. I'd be annoyed if my phone automatically logged on to a hotel WiFi (at a cost or lower performance), when the conference organiser was giving out free pass-codes. (Not the at the useless Kempinski in Berlin though, obviously - no inclusive delegate WiFi at the offload conference, ironically).
- Some applications may "come with WiFi" themselves. Skype partners with Boingo, for instance. A presentation by Sky's recently-acquired WiFi network The Cloud suggests that Sky's future video apps and content will be tightly coupled its own WiFi footprint. If I am watching Sky HD movies on my phone in a public place, I'll want to connect to its own optimised connectivity (apparently guaranteeing 1MBit/s per user) rather than someone else's that is heavily contended and which routes traffic through a video-compression box.
More generally, this fits with my concern that the telco-isation of WiFi is starting to look quite Machiavellian and unrealistic. Speaking to people with a view on the evolution of standards, some operators are apparently attempting to own and control WiFi on smartphones outright. While some level of improved control is understandable, we should be wary of the idea that an operator might control the overall WiFi connectivity on a device.
There are plenty of use-cases for WiFi which are not service-provider centric but "private"- notably enterprise connectivity, or connected-home technologies such as DLNA. If someone sends photos from their phone to their TV or home media server, that is not a "service", but merely data transferred locally over the individual's own network. You wouldn't expect an operator to be involved if you just moved the memory card, which is functionally identical to local WiFi use.
Ultimately, WiFi is a form of Wireless LAN. It's a form of Ethernet. In general, companies that don't understand LANs in general are not the right one to get wireless vesions working properly in particular. Most ethernet use is private, and WLAN is no different.
Some other points from the event:
- Offloading signalling does not appear to be well-understood yet, but was at least a topic of discussion
- There wasn't as much talk about on-device client software for offload control/management as I'd expected, a on lthough there were companies such as Roke and Onavo in attendance
- The session on Net Neutrality was lively, but didn't really touch on offload that much. The AT&T speaker was very vocal against the hard neutrality laws being mooted in the Netherlands, but conspicuously silent on how non-neutrality might impact its own femtocell traffic when carried over competing fixed/cable ISPs broadband.
- Some very good sessions on mobile broadband economics - especially around the mix of data from different devices, and the fact that for most operators, only a few cells really face congestion at the moment.
- It's worth bearing in mind that for those MNOs selling USB dongles as an alternative to fixed broadband, that means that their customers won't have home DSL/cable to which to attach a femto or use WiFi for offload....
- Offloading traffic from a MiFi-style personal hotspot (or smartphone tethering) clearly makes no sense
- There are plenty of complex connection-management scenarios to deal with around offload, for instance selecting between LTE macrocell, HSPA femtocell and various WiFi connections, especially with multi-radio capable devices and multi-tasking where certain apps have different needs.
- LTE offload is going to get tricky around managing VoIP, whether it's operator-based or third party.
- Use of WiFi when travelling internationally is going to be an important part of operator strategy. I wouldn't be surprised to see aggregate WiFi+3G roaming statistics being used to convince regulators that "average" data roaming prices are falling fast, even though the cellular portion remains very high-cost.
One other thing I'm becoming aware of: there’s quite a lot of smoke & mirrors about WiFi offload stats. In particular, a lot of the published numbers for “% of data offloaded by operator X” need to be viewed through a lens of scepticism.
In my view, WiFi usage on smartphones falls into three main categories:
- Private WiFi use, typically either in the home or office but also elsewhere. As discussed above, this is WiFi access that is used with the mindset of “having a small computer connected to a LAN or broadband” – ie those applications and content that might have been used even without having a cellular data plan anyway. One way to think about it is the type of use that you might see with an iPod Touch – which clearly isn’t offload WiFi traffic as it doesn’t have a 3G modem.
- Offload WiFi – this is the traffic which is directly moving from 3G/4G connectivity over to the WiFi access, directly substituted. This is the number that is the most important in terms of the economics – traffic which would otherwise have gone over the macro-cellular infrastructure.
- Elastic WiFi – this is linked to offloaded traffic, but represents the extra amount that users will tend to consume given faster speeds or (perceived) lower price. In other words, this is incremental and not substitutional, even if it is mobility-centric use cases (eg watching video in a café or airport)
I'll be uploading my presentation to Scribd and Slideshare soon & will post accordingly.