Pages

Pages

Monday, December 16, 2013

The beginning of the end for IMS. WebRTC is the catalyst

I've been skeptical about IMS - especially in mobile - for a long time. But while I'm wary of confirmation bias, recent events mean that I am now confident the telecoms industry is internally recognising IMS's limits. Over time its relevance will dwindle, although conservatism and forced-need will give it some residual short-term momentum.

I started writing about IMS in 2005 and published a report in mid-2006 pointing out that nobody had bothered defining what an IMS-capable handset was, and that as a result it would be at least 2010 before commercial massmarket mobile IMS services were likely. Clearly, that was an over-optimistic view, as it's now the end of 2013, and probably less than 0.5% of handsets are properly IMS-capable, and a much smaller % of customers actually use the services.

Since 2006, my stance on the technology has hardened, especially with the ongoing farce that is RCS/joyn. While basic IMS platforms make some sense for replacing 40-year old fixed networks' switches with circuit telephony-equivalent VoIP for opex saving, the same does not apply in mobile, where new and cost-optimised circuit cores are a long way from the ends of their depreciation cycles.

There is not a single service that IMS can do that other platforms cannot - typically with greater flexibility, lower cost and much faster time-to-market. Furthermore, IMS's developer base is woefully small, while telco CFOs view investing in it - at absolute best - as a necessary evil. Its current role seems to be for operators that have painted themselves into a corner on VoLTE, where they can't spare the spectrum to maintain CSFB or SV-LTE any more. A few have drunk the RCS koolaid, but even there, outsourcing or cloud platforms means that little internal investment in IMS platforms is needed for a minor experimentation or deployment. A handful of stalwarts still seem evangelical, much to the skepticism of other operators - and often many of their own staff in other departments as well.

In 2009, I light-heartedly mused whether the mobile industry had essentially "nailed the dead parrot of IMS to the perch of LTE" - ie tried to tie the standards so closely together that operators would be forced to deploy IMS anyway, despite it being a dead & moribund platform for innovation. (If you're a Monty Python fan, have a laugh at my fairly notorious blog post closely referencing the famous sketch). 

It seems I may have been prescient. This weekend marked the 4th anniversary of the first LTE launch, and yet only a handful of operators have deployed VoLTE in the wild - conspicuously not including TeliaSonera, whose 2009 network set the 4G ball rolling (Norwegian VoLTE is as rare as Norwegian Blue parrots, it seems). 

The South Koreans have it running, as does MetroPCS in some areas, although its new owners T-Mobile are more equivocal about deployment schedules elsewhere. O2 Germany has it "at a few base stations" (ie a face-saving press release) and two HK operators have made announcements of imminent launch. Only this week, UK LTE operator EE said it would "trial" VoLTE in 2014. This fits with my oft-stated summary that:

"VoLTE was the right idea but 5 years too late; RCS was a stupid idea to begin with". 

VoLTE has been plagued by myriad complexities, such as the SR-VCC handover to 2G/3G, the need for full coverage (& often small cells and fibre backhaul) to support it, innumerable integration issues, and perhaps above all the basic economic costs of the IMS platform and devices/clients needed to make it work. In many countries, voice telephony revenues are already flat or declining - which makes investment in a complex new platform, for a potentially declining service, a difficult business case to justify. Add in uncertain support from Apple and the need for a Grade-A network including indoor coverage, and the argument looks weaker still.

Even in the fixed telecoms world, there is very little non-telephony use of IMS. Network-based videoconferencing, enterprise unified comms, IM and a handful of other services are deployed but little-used. In most instances, telcos resell non-IMS third-party conferencing or UC solutions instead, while far more have partnered with Internet/app companies like Whatsapp than have launched RCS.

IMS has increasingly been made obsolete - both for applications and as a control-plane solution - as a concept by:


  • Widespread use by billions of people of the open Internet
  • Wide availability of fast fixed and mobile broadband, allowing third-party VoIP and video services to emerge at low cost, offering free/cheap services
  • User behaviour indicating that "quality of service" only matters some of the time, and that often quality is easily ignored if the price is right/zero
  • User behaviour indicating that "islands" are often seen as more valuable than interoperable standardised services, for many use cases. "Walled gardens" are fine as long as the users are themselves willingly climbing in over the walls
  • Smartphones, initially from Nokia but now Apple & Samsung have made it easy for alternative communications applications to be accessed and installed. Apple has some of its own, in particular.
  • Hopeless timescales for IMS standards development, especially for applications. The committee-led bureacracy of RCS, with lengthy monthly/annual cycles, has to compete with app players that can make decisions on new features in an afternoon, and deploy them the week after.
  • The most vibrant telecom service domains are largely outside of IMS's grasp - content, IPTV, cloud SaaS, hosting, vertical initiatives and so forth. Conversely, "session-type" communications such as telephony and SMS/IM are declining in revenue (and often, relevance).
  • The low relevance (and or focus) placed on IMS by telcos' Digital Services groups - whether they are doing Telco-OTT VoIP, content/entertainment offers, or vertical services for healthcare, transport, energy and so forth.
  • The need for telcos to accept multiple identity models rather than key everything to phone numbers and subscriptions
  • Wide use of 3rd party WiFi on mobile devices, meaning that centralised policy-control is mostly irrelevant, while operator-managed services need to work over generic connections as well as "on-net". If services are used OTT-style some of the time, then why not all of the time?

Yet despite all this, there is still a huge part of the telco and vendor community that doggedly sticks to the view that IMS's time will come. That VoLTE and RCS, and maybe some lipstick-on-a-pig API exposure layered on top, will lead to its rightful and "ubiquitous" implementation across all operators. That fully-interoperable, QoS-managed, access-integrated communications services of the future will be based upon it.

The reason for this is simple: it is standardised, and "therefore" it is the only choice. A generation or two of telecoms engineers and architects have been brainwashed into believing that standardisation is essential, innovation is done by vendors/committees, and "interoperability" at a service level is paramount - and transcends "minor" issues like user preferences, wants and needs. Now clearly, certain aspects of interop are essential - notably in the radio network to make sure that device A works with network element B - but for actual end-user services, interop is only needed for basic lowest-common denominator offers. (See here for an earlier blog post on the limits of interoperability)

In other words, the reason that IMS has been grinding on for so long - despite its limitations being obvious, and numerous "refugees" publicly denouncing it - is because the telecoms establishment can't bring itself to admit that it got the last decade so spectacularly wrong. This is understandable - it's a well-known psychological syndrome called the Endowment Effect which means you tend to like what you already have, more than is reasonable or logical. People have invested a lot of time and money in IMS - careers, even - and are reluctant to acknowledge that it's been mostly wasted. 

We also tend to have another cognitive tendency to defend our own belief systems, in spite of overwhelming evidence that they're flawed. There is even a neuroscientific reason for pushing back against the blindingly obvious, to the extent of even lashing out at it - our brains are programmed not to change complete frameworks of perception.

Which is why last week's WebRTC conference marked a turning point for me. Parts of the "establishment" are now shifting. Most notably, the term "Post-IMS world" was used by a senior operator labs representative from Portugal Telecom, as well as another from Orange showing slides with IMS potentially being an add-on to a WebRTC-centric domain, and not vice-versa. IMS is still there, but it is relegated to the same sort of "legacy interop" category as the PSTN. In other words, IMS will be about for a while, but will decline in importance, and certainly will not be where the innovation or new revenue happens. (PT has been working with Deutsche Telekom on an experimental WebRTC platform called WONDER, standing for "Webrtc interOperability tested in coNtradictive DEployment scenaRios").  

Now at the moment, these noises are coming more from the Labs and CTO offices of operators, not the architects deploying today's infrastructure. I'm sure that all those operators have IMS loyalists as well, but open dissent is often necessary before a coup. These are experiments and ideas about the future. Nevertheless, they represent an important sign that the mindset is shifting. As per the title of this post - it's the beginning of the end, not the death knell quite yet.

Perhaps more significant was a vendor speaker who is also on the 3GPP project to integrate WebRTC and IMS. Some of this was about straightforward gateway specifications & timelines, but other aspects of it seemed similar to the WebRTC-first world-view espoused by PT and Orange. There is tacit recognition that services will often be anchored primarily on the web (or enterprises' IT), but will sometimes need to interact with PSTN/IMS as a secondary role. This is a first in my view - the 3GPP acknowledging that some communications services, deployed or sold by telcos, may not be IMS-based in the future. We all know it's true - many Telco-OTT propositions are built oustide the reach of IMS - but it's rare to get "official" statements recognising it.

One other thing was around authentication and access to telcos' WebRTC services. Obviously, many vendors and some operators want to re-use the phone number, SIM credentials and so on. This makes sense for certain use cases, although the phone number remains totally inappropriate for most such instances on the web, where users may be anonymous, temporary users, "guests" and so on, rather than "subscribers". One of the possible items for the second version of the 3GPP standard is called "large-scale 3rd party access". As the event chair, I challenged the speaker to define that more clearly - I asked "Is that a euphemism for 'log on with Facebook'?". The answer? An unequivocal, one-word response: "Yes". Again, that to me is a landmark of realism from 3GPP.

The brutal truth is that many telcos will keep banging their heads on the IMS wall, because they do not have the resources or vision to see beyond standards, even if they are clunky, expensive or obsolete. Then, they will complain about more-nimble Internet competition. Some will set up Digital Services groups to build their own equivalents, or partner with Internet/app winners. But others will continue down the blind alley regardless. Having the 3GPP at least acknowledge the existence of IMS alternatives is a good start, although not enough. More clear-thinkers willing to challenge entrenched legacy viewpoints in Labs and CTO Offices is also needed - perhaps a tough task in some of the more stubborn operators like Vodafone and AT&T, which are still doggedly trying to push legacy-IMS into the 21st century.

I'd say that there is a strong argument that investors, CEOs and boards need to fire those responsible for wasting time and money on an unworkable technology. Those who have abdicated the responsibility for platform innovation to vendors and standards groups have squandered 10 years, versus the Internet players. I can't recall the last time a telecoms CTO was fired for making a poor technology choice - because usually the "choice" isn't theirs to make - it's predefined by a group of people who never talk to end-customers or developers. I still meet people saying "Ah, but you can't send a message from Facebook to LinkedIn, or a make a video call from Skype to Facetime"

My view is that IMS is reaching its zenith. It still has some brief upward momentum left - but the rockets have run out of fuel, and gravity is starting to pull it back. Whether it has reached a stable orbit, or will break up on re-entry is unclear. Some more VoLTE deployments will probably keep vendors happy in 2014 and 2015, although it is abundantly clear that it will not be the "default" telephony solution of the future. I'd be surprised if VoLTE ever gets beyond 10% penetration of the global mobile user base, given everything else that is going on with communications and voice and the moment. RCS will struggle to gain active use by 2%, I'd expect, and most likely much lower. The API argument for IMS/VoLTE/RCS is also pretty ugly, especially as IMS-capable device penetration will never get to the "ubiquity" that evangelists are promising. Again, there may be some exceptions, but the bottom line remains that IMS is providing the wrong "raw ingredients" for many use-cases, so exposing them via APIs won't help, except for some niches. It's also possible that some countries will become "IMS islands" because of local specific issues - China, perhaps.

But I expect that many operators will simply decide to avoid IMS entirely, or deploy a bare minimum for roamers and a couple of point features. Even open-source efforts like Metaswitch's Project Clearwater won't change the end-game. Lower infrastructure costs are good news, but the lack of developer enthusiasm - and the dependence on the same tired legacy interop model - is a major limiting factor. That said, helping telcos to fail faster and more cheaply is at least in line with the Internet/app ethos they'll need to learn anyway.

Overall, we're now at the beginning of the end for IMS as an application platform, while its control-layer aspects are also a poor fit with coming network realities and user behaviour. It won't go quickly, but then neither did fax or telex. 

The parrot is still nailed to the perch. But it's still dead.

(If this topic is of interest, please contact information AT disruptive-analysis DOT com for details of workshops, speaking engagements, consulting projects & published research. Or click here for details of Disruptive Analysis' WebRTC report  & update subscriptions)


16 comments:

  1. Anonymous1:58 pm

    IMS was a highly optimized architecture to re-implement the year 2000's GSM/feature phone infrastructure with IP. Modify a single detail and it all falls apart. It has so many hidden assumptions rooted in the early 2000s, it's not funny. When your service needs a user-specific P-CSCF behaviour, you are screwed. It has no good story for IP PBX integration, so it is no good fit for fixed telcos. It relies on IP addresses for authentication, so it needs a separate internet to avoid spoofing. And a tight coupling between SIP and lower layers. It should have died long ago.

    ReplyDelete
  2. Very much agree Dean. I look at a similar issue on the business case of next gen voice here in my blog post:

    http://www.alianza.com/delicate-balance-voice-without-the-investment/

    ReplyDelete
  3. "Webrtc interOperability tested in coNtradictive DEployment scenaRios" is certainly the most wondrous term I have heard recently! Amusing that this culture-of-complexity was an item in your positives column of Telcos moving in WebRTC directions :). Great disruptive article!

    ReplyDelete
  4. A key area of difference (mentioned but I don't see fully expanded here) between the open Internet and IMS is the cost, likelihood and speed of entry by innovative, and often small, new fresh development teams. The complex walls and IMS-vendor-gates keeping young hackers out of the IMS apps world are immense. Meanwhile the SoCoMo hordes - now nearly 20 million web-style developers - have immense elastic cloud power and huge quantities of open source or reasonable-cost-as-you-use tools that provably scale to hundreds of millions of users (or to 1.2 billion in Facebook's case). The expertise and learning of the first wave of Twitter, Facebook, LinkedIn, Netflix, Amazon, Salesforce and others has already been passed forward to the second wave of Instagram, Pinterest, Tumblr, Box, Jive, Workday, Uber, Twilio and many more in many market spaces/cloud-ecosystems. In each wave the consumer expertise has bled though rapidly into disruptive enterprise cloud and mobile apps. And now the third wave, where any of these established players or new innovators can easily just "add in" global state-of-the-art real-time communications is upon us - using a combination of now established communications-PaaS, raw IaaS with open source and, of course, WebRTC! This is, and also has been, an apps war where there are no sizable walls to entry and innovation on the open Internet side.

    ReplyDelete
  5. Anonymous - thanks for the input. Interestingly, IP-PBX/UC integration is part of the pitch for future 3GPP IMS/WebRTC integration, especially where the next phase possibly separates one of the elements into an "internal" and "external" one.

    Kevin - cheers, I will check it out. I expect quite a few telcos to either opt for hosted/wholesaled VoIP, or go all the way to BYOV propositions. (Or perhaps pitch 900MHz GSM as the premium telephony experience!)

    Lawrence - many thanks for the comments. And yes, I completely agree with the external Internet development momentum, which will now be able to add communications-as-a-features with WebRTC. I realised I actually didn't expand on the WebRTC catalyts theme as much as I could have done.

    ReplyDelete
  6. I agree with Lawrence that this is matter of economics. GSM overtook CDMA because of relative openness and economies of scale. The developer economics speak for WebRTC being the real winner in next wave of communications innovation over IMS. There simply are more HTML developers/companies/players than there are Java/IMS developers. The CFOs of operators are facing a decline in revenues - 3% in 2013 across Europe in general. In a stagnant business focus is on costs and scale drives them down. Obviously it's cheaper to launch new services that are web based vs. something that is sitting tightly in the core network and in the modem.

    However, I am seeing VoLTE and mobile person to person voice calls happening with IMS. Carriers opening up their IMS via Web APIs or RCS with Web APIs will not do much good for them. Interesting development and an eye opener internally, but true communication innovations are done by totally new group of players. Apps/services (beyond person to person voice calls and emergency calls) as Lawrence points out are the winners.

    ReplyDelete
  7. Anonymous4:21 pm

    Question: How VoLTE is going to be implemented without IMS? Unfortunately, that is the standard solution. CS-Fallback has performance issues. This is the situation. What are the alternatives for voice in LTE network?

    ReplyDelete
  8. Jari & Anonymous: Yes, VoLTE is designed to be based on IMS. This is another one of its critical failures.

    The only answer is that only a few operators will deploy full IMS+VoLTE.

    Others will either:
    a) Stick with CSFB
    b) Use hosted/outsourced VoLTE in some form
    c) Use a form of NGN VoIP which looks broadly similar in terms of "profile" with VoLTE but is not IMS-based, but interoperates in the network
    d) Tells customers to "bring your own VoIP" as they do not wish to offer it themselves
    e) Partners with Skype or another OTT provider to offer telephony services
    f) Develops their own Telco-OTT telephony product & interoperates in the cloud/

    The biggest myth is that VoLTE is the only form of voice service that can have QoS. This is the big IMS lie.

    If you speak to the policy management vendors or operator teams, they claim to be able to enable QoS for ANY application - hence the whole battle around Net Neutrality. Therefore if this is true, it should be entirely possible to apply QoS to any non-IMS form of VoIP as well, although perhaps with a different mechanism to VoLTE.

    Ultimately, some form of non-IMS method for LTE telephony may probably get standardised, after it is realised that VoLTE is a failure, with IMS as expensive and obsolete. By then, it will probably be too late to save the telephony revenues from the same fate as SMS.

    The stupid insistence on IMS may well have killed the cellular industry as we know it.

    ReplyDelete
  9. Anonymous6:50 pm

    Agree that IMS is overly complex. It was supposed to be a flexible architecture to deploy cutting edge services, where none got deployed.

    Having said that, the island issue in the WebRTC world does bother me (as a consumer). For one there is lock in and no customer service, even when you pay. Witness the problems with Yahoo mail, where I have been paying them $20/year for a "premium" service. E-164 address is still a universal way to reach people across islands.

    I think operators need to adopt VoLTE from a cloud provider asap.

    ReplyDelete
  10. John.9:47 pm

    Couple of comments:

    1. I don't think that anyone claims that IMS is the only way to provide voice QoS in any kind of network, indeed the PCC architecture can be (and is) deployed completely independently of any IMS elements and can provide QoS for any service that can be classified. VoLTE with IMS of course has some ways to ensure that IMS services get QoS but the LTE network could just as easily provide that QoS to a completely different service.

    2. Dean has been telling us about the death of IMS for years now and yet somehow it still lingers on with more and more deployments coming every year. No it has not lived up to the hype but if I want to deploy a large scale (millions or 10s of millions of users) voice service over IP today that provides the necessary vendor support and regulatory features (lawful intercept, 911, location etc.) then it's hard to see what alternatives there really are (just CS I think). "Roll your own" does not strike me as likely choice for most operators and partnering with a OTT is pretty unpalatable. There's definitely a momentum behind IMS that I don't see slowing (perhaps you might argue that it couldn't actually get any slower ;-)

    ReplyDelete
  11. This comment has been removed by the author.

    ReplyDelete
  12. Dean, I'm not sure if IMS is obsolete or reaching its zenith. I don't know it. I believe in certain things with certain probability. And in long term, I give IMS 90% it is going to be with us for many many years.

    Why ? there is no reasonable substitution. The rest are just "toys".

    You say: "User behaviour indicating that "islands" are often seen as more valuable than interoperable standardised services".

    My answer:"islands" are fine as long as you can relay on standardized services if you can't use your "islands".

    You say:" The stupid insistence on IMS may well have killed the cellular industry as we know it."

    My answer: "Nobody insists in IMS. let's everybody to choose what what he wants. If someone wants IMS it doesn't mean he is stupid. The combination of the reduced CS core with ICS support + VoLTE IMS covered by SR-VCC is reasonable.

    Not understood "how IMS would kill cellular industry as we know it"

    Dean, we operate live networks where people do not expect any outage and expect certain quality.
    I want to say we don't play with simple toys like Skype, Google Voice, Viber, etc.

    I work in the industry for 14 years on CS core design, PS core design, IN planning, IMS RCS VoLTE design. I like Skype, Viber, WhatsApp, I use them a lot. I.e. I'm using "islands" and I'd like them. But when thinking about future-proof platform I don't see anything better then IMS. (IMS as SaaS in 7 years from know is reasonable)

    WebRTC is cool idea and I'm very interested what it will bring to us. But we both have to admit that so far is more buzzword then something useful. It is so limited in "signalling" that you are lucky to connect A with B - when I was using that from my Chrome it was like Tin can telephone ;-)

    BTW, there is WebRTC GW for IMS already.

    ReplyDelete
  13. VoLTE Benefits
    The reasons for using VoLTE instead of separated voice/data solutions are numerous:
    1. VoLTE is simple and reliable: With its guaranteed QoS, VoLTE offers an
    operator’s subscribers a robust alternative to OTT VoIP in an easy-to-use
    mechanism. Users know that they can rely and enjoy carrier-grade services
    without disruption or uncertainty.
    2. VoLTE exceeds OTT VoIP and even 3G voice quality: Through the use of
    Adaptive Multirate Wideband (AMR-WB) codecs (12.65kbps or 23.85kpbs) and
    QoS Class Identifier (QCI) SIP signalling, VoLTE provides “HD Voice” for a
    noticeably better end user experience compared to OTT VoIP and even 3G
    3. VoLTE reduces the work to provide rich media services: As explained
    previously, in VoLTE, operators have a foundation for using RCS to enable video
    and file sharing, presence, instant messaging and enhanced phonebook services.
    An operator who has already implement VoLTE is closer to implementing RCS
    than a non-VoLTE operator.
    4. VoLTE is spectrally efficient: Because of LTE’s all-IP architecture and new
    features in 3GPP releases (such as MIMO antenna technology), voice requires
    less bandwidth in LTE spectrum than it does in 2G/3G networks. Consequently
    operators have more available data capacity in their bands for a given voice load.
    5. VoLTE reduces operator Opex: VoLTE simplifies the network by providing data
    and voice services on the same IP network, allowing operators to integrate
    network resources, optimize network and service management, and simplify
    service delivery.
    6. VoLTE accelerates evolution to LTE: With VoLTE, an operator can offer voice
    service in its LTE spectrum while harvesting its 2G/3G spectrum for re-deploying
    additional LTE bandwidth. This migration process is easier for subscribers and
    limits operator upgrade costs

    http://www.samsung.com/global/business/business-images/resource/white-paper/2013/01/SamsungCaseSmartVoLTESolution-0.pdf

    ReplyDelete
  14. Sorry Jan, just because Samsung puts a lot of marketing fluff about VoLTE in its brochures doesn't make it true!

    1) It's certainly not simple. And network QoS does not translate into either good voice experience, or overall good QoE

    2)Rubbish - AMR-WB is not as flexible or good as Opus or some other codecs. At best, it is broadly equivalent to other forms of VoIP

    3)That's a ridiculous argument - very few operators want RCS, and there's plenty of ways to implement it without doing VoLTE first. There are no worthwhile use-cases for RCS anyway, as I've written about for years.

    4) Yet to be shown how much spectrum gets freed up in real implementations. It may actually be worse because of the maths needed to deliver QoS priority for VoLTE.

    5) Nonsense, as most existing 2G/3G voice core networks are new & not fully depreciated yet. And in any case they will need to be maintained in parallel for years, if not decades.

    6) Total nonsense. If anything, the cost & complexity of VoLTE makes it harder to justify LTE deployment, not easier

    ReplyDelete
  15. I suspect the truth of the matter lies somewhere between the Samsung marketing that Jan posted and Dean's hyperbole. There is real interest in RCS worldwide amongst operators and VoLTE is proceeding into deployment - both are happening albeit slowly.

    A few responses to each point:

    1. Agree that VoLTE is not simple, but no large scale communication system is. At least it provides the features that any large operator will demand - that's certainly no guarantee of success ultimately! Agree that QoS is only one element of QoE, but it is important.

    2. AMR-WB is a huge improvement on the usual narrowband codecs in widespread use today in VoIP and traditional networks. Are there better codecs? Probably, but I suspect that the differences between WB codecs are not huge - would be interested to hear other opinions.

    3. Operators worldwide are showing a lot of interest in the RCS feature set and yes it's not the be-all and end-all.

    4. Can't speak to the Voice spectrum savings, I suspect that given the proportion of voice vs. data that this is not a particularly interesting argument.

    5. Whether or not there are opex savings will vary hugely from operator to operator. A new entrant LTE operator for example is hardly going to want to build out CS capability to handle voice.

    6. Jan's point is overblown but hardly "total nonsense" If an operator can increase or introduce voice service with VoLTE as well as a data service then that presumably will add to the business case.

    ReplyDelete
  16. Jaininder6:19 am

    Very well said.. I think nutshell is that its the time for de-linking technology & business in telecom.

    ReplyDelete