Musings on WiFi, Femtocells and the mobile Internet.pdf

[shratko1]

The rise of the mobile Internet represents enormous scaling challenges for mobile operators.  This paper takes a look at the role that Wi-Fi and Femtocells will play in meeting this challenge.  They both have a lot to offer, and the mobile Internet user experience will be greatly enhanced by leveraging their unique strengths.

Steve Hratko

SMBU Group

Cisco

[deanbubley]

To my mind, this all points to a need for much more sophisticated connection-management software in devices. There is a particular issue around multi-tasking smartphones or laptops, where there may be concurrent applications running with very different connection needs, eg optimised for WiFi, femto or macrocellular connectivity.

There will need to be a blend of both application-aware networks, and network-aware applications, mediated by intelligence in the OS/connection manager, as well as in the control layer of the carrier(s) network.

Dean Bubley

Disruptive Analysis

[shratko1]

It is hard to argue with your assessment.  The application aware network is already a reality with high-performance eGGSNs supported by dynamic policy control.  The network can now identify the application and make some suitable decision about what to do with.  Options might include blocking, rate limiting, passing it all through, etc. (peer-to-peer is an app that usually requires special treatment).  It would be even better if a mechanism existed to push this traffic off onto a less expensive network like Wi-Fi, but it isn’t clear how the network could make this happen short of dropping the user.  It has always sounded like something that is easier done by an application aware connection manager in the device.  Possibly at the request of the network.

Since Wi-Fi will do most of the heavy lifting in the mobile Internet, it might be worthwhile to draw up a list of everything that needs to happen to offer the best possible user experience in a dual-mode mobile device.  Such a list would include at a minimum:

1)      802.11u to help the mobile device to determine the capabilities of a Wi-Fi network pre-association.

2)      A solution to the Wi-Fi roaming problem.  It is certainly possible for 802.11u to advertise any roaming agreements (pre-association) which would be of great value to the user, but I think more is needed. 

3)      What is the best approach to the power control problem associated with running TWO radios at the same time. 

4)      Session persistence as the user begins to move about and loses their Wi-Fi signal.  What is really needed here?  Which applications will break and which are ok when their IP point-of-attachment changes.  Do we really need to add a lot of complexity here?

5)      What capabilities are required in our dual-mode connection manager.  Certainly an understanding of the needs of various applications running on the device.

6)      Exactly what is the role of the operator in moving users on and off Wi-Fi networks that they don’t necessarily control.

7)      And much more no doubt

One final thought.  Do operators really need to be in the Wi-Fi infrastructure business?  I’ve long been of the opinion that they don’t, but now I’m not so sure.  AT&T with almost 9 million iPhones is now a BIG believer in Wi-Fi and has accumulated significant Wi-Fi assets through their Wayport acquisition.  Is this going to be the future for all mobile operators?

[aokmians]

Steve, interesting analysis. I agree with your conclusion of WiFi and femtocells being complimentary. See my post on this subject here.

One thing to note. You article appears to make an implicit assumption that femtocell has to channel traffic via SPs core.  In fact, people are actively looking for local breakout solutions where some traffic goes from femtocell straight to the internet.  Every SP pretty much expects a roadmap of support of this feature.  The challanging aspect is the legal intercept requirements.  It is being worked on.

Given the local break out, the bottleneck would really be just the radio part.  Of course, if LTE really provides 100+ Mbps, then it is possible that the wired broadband link itself could be the bottleneck.

I also wonder if anyone could compare potential bandwidth of LTE and WiFi.  Both support 20Mhz channels after all, with LTE channel size being flexible.  If anybody saw a good study/analysis, please share.

[MikeDemler]

Great article and great discussion.

Anton, I'm not clear on the question in your last comment.  B/W of LTE vs. WiFI?  Do you mean bitrate, efficiency? With 802.11n going up to 40MHz, isn't that the bandwidth?  See this Intel article showing 802.11n up to 600Mbps.

[aokmians]

They are probably lying about 600Mbps.  80-100 is more like it in practice. At least according to published research.  It also talks about the resulting higher interferance due to channel bonding. It ain't free.  In a controlled enterprise setting it may be great, but out in the wild, it is not clear what the end result is.

I wonder if LTE would support bonding.  Maybe with a lower-powered base station it makes sense.  I guess WiFi will be the ginnie pig. 

Are there other fundamental reasons for LTE to be slower than WiFi when working with local base stations?

[deanbubley]

Some more thoughts on all of this:

1) Steve's initial assumption that the majority of data traffic from wireless devices will be in the home/office is at the mercy of future application developers, and to a lesser degree device innovators. For example in the past few months, we have seen the emergence of "augmented reality" (a sort of head-up display) using phone camera, screen & browser to overlay data on "the real world". If this gains adoption, it could massively drive outdoor mobile data traffic. Background download/cacheing of video files could do the same, and so forth, as could real-time "black-box" telemetry from vehicles.

2) It's an open question whether a mobile operator would want to offload *all* data from a particular application onto local IP breakout via femto/WiFi, and whether that was controlled in the network, in device hardware or in the specific user application. For regulatory reasons & to collect valuable subscriber information, a hybrid approach might be desirable. So, an operator might want the *metadata* that I'm about to download a 2GB movie from Netflix, its title, age-suitability etc. So that (small) data volume has high value and should be captured in its data centre somehow. But the bulk of the media traffic might be safely offloaded - although potentially the operator might want to do its own advert-insertion, which offload could make problematic. There may well be offload vs. business-model compromises to be made.

3) I am currently investigating whether fixed/cable operators might be able to offer "managed intelligent offload" services to their mobile SP counterparts. While some operators like Orange or AT&T clearly have their own DSL customers, the majority of households will not be quad-play/single-MNO, so there will often be a patchwork of ISPs or fixed-line providers involved in femto/WiFi management. There is a portfolio of possible offload service I can think of, from prioritisation/optimisation, through to device management, metadata capture, lawful intercept "by proxy", connectivity to CDNs or other networks, multi-MNO national roaming etc.I can see both femto-centric and WiFi-centric options here.

I'd love to talk to anyone with views on the practicality and commercials of managed offload as a new business model.

4) Some people in the cellular / femto community seem to view femtos as a substitute for WiFi for in-home wireless access. Personally I disagree, but equally I do not believe WiFi penetration will become ubiquitous in mobile phones - I see a ceiling of maybe 30% globally, or perhaps 50% in North America & parts of Europe.

5) It's entirely unclear how any of this might work for the 70%+ of the world's mobile users who use prepaid services rather than monthly subscriptions.

6) There's a range of national-specific things that may impact the WiFi/femto balance for offload, such as existing prevalence & use of WiFi, laws on net neutrality (eg can the DSL company block traffic destined for femto gateways), whether handsets tend to be operator-provided or open-market (affects software pre-load & configuration) and 100 others.

7) In some early femto rollouts (eg Vodafone HSPA), the real bottleneck is upstream DSL/cable bandwidth.