Speaking Engagements & Private Workshops - Get Dean Bubley to present or chair your event

Need an experienced, provocative & influential telecoms keynote speaker, moderator/chair or workshop facilitator?
To discuss Dean Bubley's appearance at a specific event, contact information AT disruptive-analysis DOT com

Friday, December 04, 2009

LTE Voice: the more I've looked at Circuit fallback, the worse it's looked

I've been talking about the problems with voice on LTE for more than two years now. I first published a report on VoIPo3G in November 2007 in which I discussed in great detail the problems with IMS voice, and the fact that there was no simple, standardised version of "basic VoIP" suitable for operator deployment. I'd previously written about the issues with IMS-capable handsets in 2006, and VoWLAN the year before. More generally, I've been covering wireless VoIP since about 2001.

A common strand in my analysis has been that the technical standards are often completely divorced from considerations of end-user behaviour and experience, the needs of application developers, the practicalities of handset design and the realities of business models.

Like politicians, there's instead usually a focus on control and ideology, rather than pragmatism.

That's not to say that every new technology development has to start with some sort of fluffy "inclusive" focus group approach, or that pseudo-cartels should invent things that competition authorities will frown upon.

It often just means that more consideration needs to be paid to questions like:
  • "Hang on a minute - how's this actually going to look in the hands of the user?" or
  • "Isn't this going to make it worse than the older solution people have already? Who's going to buy that?" or
  • "Isn't that going to break all the other applications running on the phone? What happens if it's multi-tasking?"

A very good example of this is one of the proposed standards for providing voice over LTE networks, Circuit Switched Fall Back (CSFB). This is the 3GPP approach to supporting voice which drops the LTE connection back to 2G or 3G to make or recieve a phone call. It is generally positioned as an "interim" solution before moving to full IMS voice.

I've written before about CSFB, and it's also been discussed in many of the comments threads on my previous posts on LTE voice, such as here, here, here and here - many of which also cover ongoing problems with IMS voice. (The recent OneVoice announcement is a possible medium-term solution to some of those problems).

Some of the issues that keep cropping up include the latency involved in the LTE-to-3G/2G process, the impacts on any data applications running on LTE at the time of a call, and the need to have overlapping coverage of older networks everywhere you put LTE.

I've now done a more thorough analysis of CS Fallback's flaws. I've come to the conclusion that it's not just awkward, it's actually terrible - worse than useless, in its current incarnation at least. Going back to my questions before, it seems clear that nobody ever said something like:

"Hang on a minute, LTE phones will be expensive, so our best customers will buy them first.... but it will give them a worse telephony experience with CSFB than their existing handset. That's never going to fly!."

(Incidentally, coming back to a discussion from another thread, one commenter asserted that fallback could be achieved in about one second. I cited 2-4secs, based on a 3GPP submission I'd seen, and we agreed to disagree in the absence of hard evidence. I spoke to a *very* senior person in handset RF development for a major device vendor earlier in the week, who has more cellular patents than I've had hot dinners. His estimate was for 6-12 seconds extra latency. For an LTE-to-LTE call or SMS, it would be quicker to use Morse code).

I've now written a white paper covering CSFB's major flaws, which actually seem pretty extensive even beyond what I've written above. It almost looks like it was designed to make IMS voice look good by comparison.

The paper is available for download from here. But first, some disclosure. It has been commissioned by Kineto Wireless, the chief proponents of the main alternative VoLGA, which has evolved from UMA/GAN. In common with all Disruptive Analysis' sponsored material, I only take on topics where I already have a strong opinion - I'd written positively about VoLGA since March , despite having been a thorn in UMA's side since its launch in 2004. (I'm convinced that somewhere in Kineto HQ is a dartboard with a picture of my face on it). It's not perfect either, but it seems much better than CSFB or IMS.

I'd also previously written about CS over HSPA (which is the same basic principle of 2G voice over a 3G IP bearer) in early 2008. I'd even originally suggested doing "2G over 3G" as a possible better use case for UMA than WiFi, way back in 2005/6, albeit a bit flippantly at the time.

The bottom line is that I think that 3GPP should reconsider VoLGA or something close to it. CS Fallback looks like a terrible interim solution for voice-on-LTE, especially for operators who aren't sure that their particular endpoint is IMS voice. The current two-solution approach (CSFB / IMS) seems guaranteed to either promote 3rd-party VoIP solutions, or delay LTE entirely. If the whole thing *has* been some sort of standards conspiracy to force IMS into the hands of the unwilling, I'd make a suggestion to include some game theorists in the discussion next time.

One thing I should note for completeness is that there are a couple of other options out there, notably NSN's Fast Track and another from Acme Packet and Mavenir. I haven't had a chance to drill into those as much, but from an external perspective they don't seem to have as much traction or behind-the-scenes support as VoLGA.

Once again, the white paper on why I feel CSFB is "not fit for purpose" is here.


Kevin Mitchell said...

Dean, I'm shocked. UMA technology has been adopted by 5-10 carriers worldwide. Compare that to over 100 for IMS (sure, most are fixed, but mobile are laggards to SIP/next gen voice period). Major tier 1s MNOs have already accepted IMS as their VoLTE long term solution (generally as a converged core for fixed and mobile with fixed apps leading any mobile SIP). While Acme Packet can participate in the 3 approaches (VoLGA, MSC VoIP, IMS or SIP-based variation) to VoLTE , but obviously, we're advocates of SIP. For an interim approach, I like the MSC VoIP approach much better (service parity + next gen signaling + SIP-based presence, IM options). IMS has some negative sheen on it currently given the complexity and delays to roll-out (and your lack of love for it), but I have no doubt that SIP is the way forward. VoLTE as a service won't be needed until 2013 (although architectural decisions will be made well in advance of that) given handset availability and my own doubts about the timing of LTE services. MNOs should launch SIP communication clients with their LTE "data only" services (most can generally use their "fixed VoIP core" to do this) so to keep that subscriber's communication services on net vs losing future business to over the top.
There could be a quasi-FMC opportunity, bringing together the communications for the mobile device and the LTE connected laptop (multi-device environment with CS voice to 2G/3G mobile and PS voice to SIP client on LTE laptop).

Dean Bubley said...

Hi Kevin

Thanks for the feedback.

There's four ways of doing voice over LTE

1) having a SIP VoIP client on the phone connected to SIP/IMS VoIP server in the network
2) having a CS client on the phone connected to an MSC in the network
3) having a SIP VoIP client on the phone connected to a SIP gateway in the network which sits in front of the MSC.
4) Using some form of completely separate SIP client & server, or other protocols like Skype.

Overlaid on this, there are three nasty areas of complexity and cost:

1) Swapping out MSC and CS core for IMS core
2) Getting handsets with usable telephony clients, either IMS/VoIP or some "plumbing" protocols at a lower level
3) Not creating "collateral damage" to other aspects of the user experience

The IMS Voice option hits all 3 problem areas. My concerns with IMS handsets in particular are long-standing.

CSFB and VoLGA obviously bypass #1. They both need some work down in the device firmware to either do the LTE-3G switch, or do the CS-over-IP tunnelling. However, CSFB appears to fall foul of problem #3 in a set of major ways.

Option 3, using an MSC/VoIP approach, makes the network side easier, but still requires all the handset-based VoIP/SIP/IMS horribleness to be confronted, especially if it's going to be optimised to the radio netwiork rather just a "vanilla SIP" client.

That said, I do agree that operators should put their own SIP-VoIP clients on devices or licence them from existing Internet players. But that's for a *secondary* telephony software, a bit like having Skype or other 3rd party clients on PC or iPhone. Even by 2013, I don't think SIP VoIP will be ready to be the basis of *primary* mobile telephony on normal LTE phones


Dean Bubley said...

Obviously both CSFB and VoLGA have network-side changes as well. But even though WiFi UMA's not been popular for various reasons I completely concur with (and predicted), I think the UMA/GAN security gateway bit is now a relatively low risk element.

Kevin Mitchell said...

Dean, these SPs are investing in VoIP/IMS regardless (it has started in their fixed division sides). It's already started and in some cases, fairly substantial. All scenarios deal with problem #2: client readiness (or in distribution). Admittedly, the client side for mobile IMS/SIP has been slow going, but I think progress is being made.

I'm curious why the VoLGA group want to spend their time attacking CSFB (which has little backing that I am aware of) and not IMS. One "major backer" has already pulled out of VoLGA.

You cite relatively broad support (not all of it public) for VoLGA. We just don't see it. Relatively broad is still less than 100 projects and a massive dedicated vendor ecosystem to SIP.

Let's meet again here in Dec 2013 and see where the extent of deployment of LTE services and handsets as well as readiness of SIP to be the next generation signaling protocol for all voice.

P.S. Why does VoLGA require IPsec? It's not the untrusted Internet/ anybody's WiFi access network of UMA FMC. This would run over licensed, managed, secure LTE access networks. Why the encryption requirement?

Dean Bubley said...


The more I think about it, the more fixed-IMS looks like it can "get away with" much, much more than mobile.

- for fixed IMS, there's usually an operator-managed box like a home gateway, with the actual phone as a "dumb endpoint" that may or may not have SIP or just plug into an analogue port
- people don't really expect a fixed phone to have a "UI" and certainly not do things like multi-tasking apps
- in fixed broadband, there's no issues about patchy coverage, there's plenty of bandwidth, low latency and a single connection, so you don't need to mess about with many optimisations, worry about handovers etc.
- you can plug a fixed phone into the wall socket or use power over ethernet, so battery life isn't an issue. Especially for things like presence with keep-alives.
- no legacy messaging clients or phonebook contact lists to integrate
- no appstores or open-OS's for fixed phones so Skype can do it better on the same device

(Slightly different in the enterprise, but that hasn't exactly been a roaring success for hosted PBXs and whatnot either).

I'm moving to the view that 3GPP needs to tear up the IMS spec for mobile and start again with a better idea of *what customers want* and *what devices can do*. And it needs to assume a broad range of business models that may *or may not* include traditional subscriptions and an end-to-end control plane.

The "client side" has been so glacially-slow, that the rest of the device industry has long out-accelerated it. The problem space is now no longer even the right set of questions. Ask Vodafone why they're not using RCS for 360.

Good question about IPsec though. Probably expedient so that as much UMA/GAN stuff as possible could be re-purposed quickly.

I suspect the VoLGA group doesn't need to attack IMS, because they can leave it to people like me. Plus, mobile people seem to like to say that IMS is their "long term goal" because it saves them from having to confront uncomfortable realities too soon.


Gabriel Brown said...

Hi Dean -- how much progress do you estimate VoLGA has made in making VoLGA clients mandatory (or at least likely) on LTE devices?

Thanks, Gabe

Dean Bubley said...


I think there are a few steps before VoLGA clients become common / mandatory on LTE devices.

There needs to be some clear and growing operator interest, plus some additional motion in the standards world. This will then persuade device vendors and HW platform vendors to start to ensure that it is possible as an *option* in the future.

There may well be a few prototype or pilot products at that stage. It may be possible to hack about with Android or the new Symbian to produce a pre-standard implementation. Or there might be a keen vendor or two who'll try a special device for trials & gaining experience.

The next stage is for certain operators to start defining hard requirements, which makes ROI calculations easier if the OEM guys can start to balance development costs vs. incremental revenue.

There will also be a parallel push to get the functionality into the large chipset vendors' offerings.

It's a bit like most network-type enhancements to phones - doesn't happen overnight but through a progressive series of actions.


Anonymous said...

VoLGA loses a supporter: http://www.unstrung.com/document.asp?doc_id=185481&

Dean Bubley said...

Possible to interpret Ericsson abandoning VoLGA in a number of ways.

One is that they didn't think they'd make enough money from it on a standalone basis - either because of limited adoption, or because of lack of competitiveness/differentiation against other vendors.

Another is that they see VoLGA as an indirect threat to mobile adoption of IMS, and have done an analysis of their potential longer-term revenue streams with/without the popularity of VoLGA.

The last option is that they think LTE adoption will be much slower and patchier, and mostly used for data in-fill rather than voice. In that scenario they'd probably just toe the 3GPP line on IMS-VoIP and CSFB rather than invest more in additional solutions.


Davide said...

Sorry guys, I know this is a work-related blog, but this is too good and related to the topic to not put it in here.


Iain said...

> the whole thing *has* been some sort of standards conspiracy to force IMS into the hands of the unwilling

To which the obvious answer is:
"Never attribute to malice that which can be adequately explained by stupidity"