Pages

Pages

Sunday, June 20, 2010

A counterpoint to femtocells - are they really necessary?

I've just read Andy Abramson's long and impassioned blog post about the pointlessness of femtos.

As he points out, I'm due to meet him for a few drinks later on this afternoon, so I'm sure we'll drill down a bit more. In part, I agree with him - WiFi is much better for *certain* applications and use cases.

However, I'll make a few comments upfront.

Firstly, WiFi can be a royal pain to set up, especially on smartphones - and *particularly* where it involves a third-party login. It's also difficult to monitor and control - and expensive to support if someone calls in for assistance. For this reason, many mobile carriers are wary of trusting WiFi as a suitable mechanism for either coverage extension or capacity offload.

As an example: while I've got my iPhone set up for my home WiFi's security, I hadn't provisioned it for BT Openzone - which (in theory) I should get free access to, courtesy of Vodafone - which would benefit from the offload and coverage extension when I'm in places like Starbucks. Result - 30 mins of frustration trying to use the 3G connection & browser to log into Vodafone's personal account and self-care page. Firstly the usual forgotten-password thing. Then hunting around the self-care system trying to find "switch on support for BT Openzone". Then Google the question, and be told "First, log into your Vodafone account". Aaargh. I've given up. Maybe I'll bother trying it from a PC at some point. Or seeing if it's driven by the Openzone splash page.

Or maybe Vodafone will just put femtos in all the UK branches of Starbucks, and I won't have to think about it.

Yes, I'm sure I can get some super WiFi log-on app thing for my iPhone which would automate it all, but frankly I can't be bothered to go looking for that either. I'll just use the 3G, even though I'm sitting in a basement on the edge of coverage.

So.... point #1 for femtocells is the pain / laziness / ignorance factor around setting up WiFi on devices.

Point #2 is the difficulty of extension of operators' own in-house voice services over WiFi. Yes, it's possible with UMA or some of the earlier VCC-style SIP approaches. However, those are quite expensive and complex to deploy, test, maintain and support - not to mention having similar issues and dependencies on the connection manager as discussed above. My views on UMA have been pretty negative since 2004, so it's fair to assume that's an argument I don't need to re-hash - although at least it now (finally) supports 3G.

Now clearly, many of Andy's clients (and indeed Disruptive Analysis') are quite happy when operators cannot extend their in-house voice services, as they are providing alternatives. I'm also a fan of Skype, Truphone, fring and so on... but I also recognise that many customers still prefer their existing mobile operator's own-brand voice service, or cannot be bothered to shop around. There are also issues about porting mobile-specific numbers (especially outside the US) to VoIP providers.

So - point #2 for femtocells is provision of ordinary circuit mobile voice indoors at good quality: which is desirable for many users, and most carriers. On the other hand, as Vodafone found out this week to its cost, this only works where the user has a 3G phone and decent-quality fixed broadband. (Most femtos don't support 2G GSM - although the CDMA variants support 1x).

Point #3 is around regulatory oversight and the perception of mobile operators that they are being held to a higher standard of content/access/application control than fixed operators - especially in markets with anonymous prepay. Access via a femto takes the data traffic back to the core network, where it can have whatever policies the operator desires applied to it, in the same way as macrocellular traffic. If the fixed broadband is the operator's own, then clearly this doesn't apply as it has a policy enforcement point anyway - or if the fixed provider has enabled a "managed offload" service of the type described in my recent Broadband Business Models report.

A related opportunity, point #4 - also discussed in that report - is for femto data to be *faster* than WiFi (or prioritised), if the mobile operator cuts a deal with the fixed/cable provider. Let's say you've got an 8MB ADSL service. But the line is capable of supporting 20MB. In theory, the mobile operator could arrange for your femto to deliver *more than 8MB*, essentially turbo-charging it vs. your own WiFi, if they are prepared to pick up part of the bill, or pay the fixed-line telco a "two-sided, slice and dice" fee. That's very much a "happy pipe" strategy of the type I've been mentioning recently.

Other femto rationales:

  • Open-access femtos offloading 3G data traffic without your need to have any WiFi setup or switching
  • Femtozone applications and services
  • Macro (outdoor) femtos providing extra 3G capacity in hotspot areas
  • Managed connectivity & security for home-workers from their enterprise. Yes, this is do-able via WiFi as well, but I can see this being linked to prioritisation and other corporate-grade teleworking services, such as operator-provided cloud computing resources.
But the final one is operational. One thing I've observed recently is that many radio network planners and engineers are deeply, inherently conservative about their infrastructure. While many of them would prefer to maintain the "outside-in always wins" approach, building macro base stations and towers, there is an acceptance that this might have to change in some circumstances. In those instances, femtocells have a bit more control - and are a bit tighter-integrated into the mainstream network operations and planning - than WiFi. They use licenced spectrum, standardised security - and are likely to be provided through the same vendors and integrators. They are much more likely to *trust* femtocells than WiFi provided via unknown parties. Of course some operators - AT&T for example - seem to have grasped the nettle earlier than others. But do not underestimate the amount of WiFi-skepticism left in the RAN community. Femtos aren't seen as that much better - but at least in "greater femto" format (or picocells), they're closer to the real thing.

Incidentally - I'll be at the London Femtocell conference on Thursday, and the Femto Forum awards dinner on Wednesday. I'm also based 3 mins round the corner from the Landmark Hotel, so if anyone wants an off-conference meeting (or a beer) let me know.

NEW Mobile Broadband Traffic Management Paper

NEW Broadband Business Models Strategy Report

8 comments:

  1. John.9:54 pm

    You missed one - cost. Femtos are expensive, either the operator provides them for free and sucks this up or tries to flog them or has some kind of subsidy in place. I just can't see any kind of take up unless the price gets near $0 (and if/when that happens you have all of the problems with spectrum management that you touch on - these are a huge impediment to rolling femtos out in large numbers IMO).

    Dead duck I reckon, there are niche applications certainly but there is no way we will all have these in our houses any time soon.

    ReplyDelete
  2. John - Softbank is already providing femtocells for free in Japan. With a free fixed broadband connection in some cases too.

    Now yes, Japan is often an exception rather than an example, but with PicoChip reckoning that it's got reference designs down below the $100 mark, I don't think we're too far off the point at which it offsets incremental capex.

    O2 UK has what, 2 million iPhone users? And it's been bumped into an extra (I think) £300m capex plan for extra capacity? Assume half the usage is in venues with decent broadband, and it would be cheaper to put in femtos....

    ReplyDelete
  3. Exactly - 100 smackers per subscriber?! How many operators can afford to absorb that?

    By comparison, a WiFi DSL CPE will be in the region of $40 or less. A DSL only CPE can be as little as $15.

    I don't think there's any way to usefully comment on your O2 example unless you happen to be employed in their network planning dept.
    Where exactly do they require extra capacity?
    What are the additional spectrum management/acquisition costs and implications of using femto to address these requirements?
    How much will dense femto deployment offset these additional requirements?

    etc etc etc

    ReplyDelete
  4. Yes, agreed that $100 is definitely significant, especially for lower-end users. I suspect that's the reason why much of the initial focus has been on marketing to groups such as:

    - Small businesses
    - Multi-user households (eg $100 between 4 users is less horrible, although I'm not a huge believer in family plans)
    - People who are genuinely likely to churn ($100 not so bad as a retention cost)
    - People in particular areas where the network gets congested & it's difficult to locate new cell sites or add extra carriers to the current ones.
    - "Professional" or enthusiast mobile users who are prepared to pay for gadgets. Anyone dropping $700 on an iPad is surely enough of a mug to stump up another $100 for a new and exclusive femtocell....

    But those niches should be enough to take volumes from the current 1 million to perhaps 10 million over the next couple of years. Then let's see what the price points look as scale curves kick in.

    O2 - haven't got the details on network planning either, and your points are well made. But the point I was making was that if operators are thinking about incremental capex in the broad range of a couple of $100 per user, then we're at least starting to get into the right ballpark.

    I'm not expecting to see femtos flood the market - but I can imagine the targettable niches growing faster.

    On the DSL CPE cost - yes, it might be $40. But then add in another $50 when Mr Non-Geek calls up customer support to inquire about where he finds his WPA2 on the Linksys box his nephew bought him in 2004.

    ReplyDelete
  5. There are plenty of good use cases for femtos as you say but these are still niche and it is perhaps the scale of this niche that we are discussing. I suppose 10m units over a couple of years is a possibility if operators continue to throw money around by giving these away but otherwise? If these devices didn't have severe technical (spectrum management) implications I could just about believe that a critical mass could build up and carry them forward but the fact is that the more of these that are deployed the bigger the headache.

    I'm not sure I understand your WPA2 point. Surely any hi-tech product is subject to the same kind of obsolescence problems? Or is it that you imagine femto cells will be completely P&P? How much does the support call cost with Mr. Non-geek who has to twiddle a NAT or UPnP setting on his no-name CPE to get his femto to work?

    ReplyDelete
  6. Anonymous3:10 am

    We should be careful in comparing the cost of a non managed unlicensed WiFi device with a Femto which provides the operator the ability to manage remotely, determine and resolve interference issues, tamper detect, remote upgrade etc. These features are enabled in Enterprise WiFi APs which retail for closer to $300.

    ReplyDelete
  7. John.7:41 am

    Anon- the point is not that these devices are directly comparable but rather that these are the kind of prices that operators pay for CPE. In fact, a $40 WiFi DSL CPE will also include 1 or 2 lines of VoIP ATA functionality and bog standard WiFi/DSL gateways will be even cheaper ($30-$35 in volume). Whatever way you slice it, $100 per CPE is very expensive and I imagine it's very hard to make a viable business case based on such a high price.

    As an aside, operators are increasingly deploying fully managed DSL gateways with TR-069 capability.

    ReplyDelete
  8. Strider5:13 pm

    I don't understand why anyone would pay for a femtocell. In fact, carrier should pay the user for installing one in their home for several reasons: a) improving coverages in cases where it's poor -- basically providing the service that the user is already paying for, and b) for off-loading traffic from the expensive wireless networks.

    ReplyDelete