My IMS=Dead Parrot post the other week drew a significant amount of attention and argument, as I anticipated. It was followed by more comments in another post.
One of the consistent arguments that was advanced was that IMS is the only standardised mechanism for the control plane for telco-grade IP networks.
From my stance, a central part of my argument is that I don't see IMS as a suitable platform for future revenue-generating, customer-demanded services and applications, at a device level or through integration with Web/consumer electronics/enterprise/IT. I also think it is inflexible in terms of supported business models, for example around wholesale or SIM-free adhoc services.
In a nutshell therefore - is it better to have:
a) A cast-iron control layer for services that nobody wants, or...
b) A less-optimal control layer that permits more service innovation?
From a million-foot view, are we moving to a situation where the notion of a single, all-encompassing telecom control plane is a ridiculous notion, similar to the idea of a government controlling all its citizens through a single database and monitoring system?
(See what I just did there, political wonks reading this?)
Are the costs and inflexibilities and "injustices" (in technology, anything which is user-unfriendly) worth it, just to achieve pure elegance of the ultimate machine?
The more I think about this, the more the whole notion of a single IMS control plane for IP looks exactly like a totalitarian state, where everything is done "for your own good". Policy control is the equivalent of the nannying Health and Safety Executive, the HSS and SIM is the equivalent of the National ID Database and ID cards, DPI is the equivalent of pervasive CCTV cameras. And the whole thing is hugely expensive for taxpayers (end users) and doesn't actually work because there are always flaws in the system.
Personally, I much prefer to live in a libertarian society, where there are specific checks and balances at particular points. I'm happy with passport controls at borders, or fines for people jumping red lights. I don't mind tax evaders being pursued. I'm in favour of jury trials and punishing prison sentences. But I don't want to live inside The Matrix.
The Internet works well enough without a centralised control plane. Various point solutions like CDNs or peering points act as a form of decentralised, distributed and collaborative control, which works pretty well, most of the time. Other large-scale systems work well without central control as well - in fact, the whole basis of capitalist economies revolves around the concept.
Apologies for the high-level "philosophical" nature of this post. But after the last week or two of discussions, I am increasingly of the opinion that IMS has a flawed central assumption: that an all-encompassing "control plane" is necessary, feasible or even desirable.
Dean Bubley's Disruptive Wireless: Thought-leading wireless industry analysis
OK OK, so you have a problem with authority.
ReplyDeleteAssuming that operators will want to transition all of their systems to IP over the coming decades, how do you suppose a call will be set-up from one network to another?
Dean, your libertarian statement is good maybe in the UK and the USA, but what about the rest of the world were people actually prefer some control of the state.. Your view is very reduced, very typical American/UK and very non-applicable in continental Europe (not even for right extremist I guess)..
ReplyDeleteIn your libertarian state, is possible that some evil people (bankers, brokers, and other idiots) take advantage of the system to get richer and richer, making the rest of the world poorer and poorer and create a financial crisis like the current.. than later everyone else has to pay for
(i do not see something similar happening with your libertarian telco scenario but this was a very easy demagogic comparison as many of the ones you do)
The SIM or SIM-free issue is in my opinion pointless, I am ok with everyone having to identify for certain things (for example having a weapon, buying alcohol drinks or other drugs, gambling in a casino)
That internet works well, is your humble opinion.. Many people would disagree, the whole music industry for example, the people who have been victims of scams, people whose privacy has been violated, people whose honor has been offended, parents who would like to have their children getting the benefits of the Internet without the current risks of abuse, etc!
I also agree that IMS is too much of a complicated system which solves many problems that none has yet, and thats why i do not see a point for it on its current definition (maybe FMC or fixed operators IP services) by no mean i see it as an alternative to plain Internet i think if there is an IMS deployment it has to integrate Internet services
But you have to admit there are some situations, for example, Emergency calls that i do not see how that would work in your libertarian state... Even libertarians like you would agree that mobile phones are very useful in emergency situations and without identification, location, a standard system supporting it and using Skype i do not know how that would work..
Anyway, your view is for me very limited and very extremist (very demagogic right winded party view), it would enrich your blog if you made an effort to understand the views of other people and put a scale of grays between your black or white
Anon: "I also agree that IMS is too much of a complicated system which solves many problems that none has yet, and thats why i do not see a point for it on its current definition (maybe FMC or fixed operators IP services) by no mean i see it as an alternative to plain Internet i think if there is an IMS deployment it has to integrate Internet services "
ReplyDeleteI think this notion that IMS is over-complicated is exaggerated. IMS is no more complicated than 2G/3G or PES-NGN and IN or even the old CS PSTN. The fact is that these massive communication systems are complex, IMS also needs a certain complexity but I don't think that it is any more so than existing systems.
It is also not necessary to implement everything at once, the core IMS for example may be sufficient for a lot of operators.
Anon1: No, I have no problem with authority. I have a problem with authoritarianism.
ReplyDeleteTo continue the politicspeak: "The central characteristic of an *illiberal* democracy is that institutional political processes are skewed in favor of the incumbent regime"
How do I suppose calls will be set-up from one network to another? First, you have to define "network" in the context of likely future communications modes. (And "call" although I expect conventional telephony to endure). I'm expecting a mix of (generally large) islands and various points of inter-operation. There will be a mix of e164 numbers and other identifiers.
As Moore's Law proceeds, it should become cheaper/faster to have better interconnection points, even if the island/border architecture is less elegant or controlled from an engineering perspective.
Anon2: First time I've been called an extreme right-wing. I didn't think there was too much provocative in my comments to be honest, although absolutely I'm using the UK as an example as I live here. We suffer from the type of soft authoritarianism I describe above, which makes me well-suited to seeing parallels elsewhere.
I also steered very clear of areas like stiffer financial regulation, which in fact I'm broadly supportive of, but which is off-topic even in a blog post like this.
Actually I'm more attuned to organisations that content that the traditional left/right polarised split is irrelevant and (literally) one-dimensional.
It strikes me that emergency calling may be a strong argument for retaining a basic-level circuit switched infrastructure, sometimes separate, and sometimes as an overlay on IP (eg perhaps via VoLGA for LTE, for example). Having a separate or parallel system reduces the risk of a single catastrophic failure mode, I would have thought.
While I am obviously not supportive of criminal activities on line, I am also not supportive of detailed scrutiny, as I often don't trust the scrutineers. The Internet clearly "works" in the sense that it has changed the world (predominantly for good) over the last 10 years in particular.
I don't often introduce politics into the blog for obvious reasons - I run a business, not a lobbying organisation. But in this specific instance I saw some interesting parallels & analogies which I thought would help clarify things like IMS.
I find it genuinely interetsing (and not discussed anywhere else, I think) that some of the most IMS-friendly countries include some of the most traditionally "left-wing" politically, such as France and Sweden. It's possibly a coincidence, but also possibly not.
DB: "How do I suppose calls will be set-up from one network to another? First, you have to define "network" in the context of likely future communications modes. (And "call" although I expect conventional telephony to endure). I'm expecting a mix of (generally large) islands and various points of inter-operation. There will be a mix of e164 numbers and other identifiers."
ReplyDeleteI'm struggling to understand this. Do we agree that the communications networks of the future (fixed and mobile) will all be IP based (transitioning over time from CS). If so, how will a user setup a call from his Vodafone end-point to his friend's Orange end-point (could be WiMAX, LTE, Fixed, CS, Google, MSN etc.)?
i.e. What mechanism will control the call routing, any associated services, regulatory functions, billing if required, etc etc?
IMS is one solution, certainly not perfect but a solution nonetheless, you have not proposed anything beyond vague statements about interconnection and Moore's law, can you expand on your vision?
It's likely to be a mix of solutions, plus gateways.
ReplyDeleteWe will clearly have to deal with a messy, heterogeneous world of telephony, as we do in most other areas of technology. There will certainly not be just one type of VoIP architecture - not IMS, not Google, not Skype, not XYZ others, but a blend.
Yes, there will need to be gateways, transcoders, address resolution, APIs and so forth. Some features or controls will not map perfectly or ubiquitously.
To your specific example - it will also depend on whether the "Orange" user always uses the "Orange" voice service. It's perfectly possible - even today, ironically using fixed IMS - to have a Vodafone softphone running on a PC connected via Orange DSL (or a Voda femtocell, come to that).
I certainly see that as we move to IP, there will need to be much more justification and customer value, in linking access and voice service from the same provider.
So rather than "which mechanism" I think the better question is "which mechanisms" (plural). It will vary widely, with many companies benefiting from providing the necessary glue and intelligence to converge - or differentiate - betwen them
DB: "It's likely to be a mix of solutions, plus gateways."
ReplyDeleteI should have been clearer, let's assume Vodafone and Orange want to continue providing voice services while migrating their networks to IP. What mechanism will they use to allow their users to call each other and for users in one network to call users in the other?
I accept that there will be lots of other IP-based voice applications available to users (Skype, Gtalk etc.) but it's reasonable to assume that V & O will also want to continue to offer these services and not become a simple bit-pipe.
Now, if there is an alternative architecture available that will satisfy the requirements of these huge operators, what is it? where is it? Perhaps it is yet to emerge?
Anon: "I should have been clearer, let's assume Vodafone and Orange want to continue providing voice services while migrating their networks to IP. What mechanism will they use to allow their users to call each other and for users in one network to call users in the other?"
ReplyDeleteGood question, I'm not sure of all the various current/proposed solutions for network-to-network interop. How does it work today if I have a BT fixed-VoIP phone and I call someone in France on Orange VoIP?
Bit out of my focus area to be honest, but from the past I seem to remember the likely options are direct A-B VoIP peering connections (common between large operators I seem to recall), a central clearinghouse approach like the GSMA's IPX, a third-party one like XConnect, or probably something Internet-based using ENUM, which I've never fully got my head around.
Or I guess you could use the rump of the PSTN or even one of the Internet services as an intermediate transit platform.
In mobile you've obviously got a few other issues to content with like roaming, but that could perhaps be done as an exception, rather than something which drives the shape of the whole architecture. No point in the tail wagging the dog.
I've just had my attention drawn to this post:
ReplyDeletehttp://www.convergenceconversation.com/posts/mike.kiely/bt-21cn-re-focus-and-ofcom
by Mike Kiely from the Broadband Britain Group, which appears to be a lobbying/advocacy group so is probably not entirely unbiased.
According to that article, BT has re-designed its 21CN network *without* an IMS-type control layer, but with more focus on transport and high-speed access.
First time I'd heard it articulated that way, I have to admit - but I'd be interested in anyone familiar with 21CN commenting on this.
Where does it say that BT have dropped IMS? It seems to imply that they are not pursuing AGCF control of MSANs for PSTN emulation but I don't see any mention of IMS being dropped altogether?
ReplyDeleteHello,
ReplyDeleteI won't elaborate on IMS as a centralized, state-controlled system to rule them all, I think it is clearly FUD.
I will just say that I can't understand you when you say that IMS does not allow the emergence of innovative services. Of course, as an analyst, you are fueled by the "vision" of operators and classical telco providers which is very narrow-minded. But I think there is a place for a bridge between web services and more telco features. I have worked on half a dozen really neat apps mixing social networking, location-based services, Web API's, Flash-based communication platforms... and all these projects were integrated in an IMS environment. They could use Diameter to send a variety of billing messages, that could provide service providers with the ability to set up any model to bill for the service.
But I must confess that the economic downturn has put an end to most of these experiments. Now the priority is given to projects Telco are comfortable with, because they fit their habits. But you shouldn't blame the IMS for it.
And IMHO, "solutions" to IMS issues proposed for instance by VolGA are far less open and innovative than what IMS could offer for service providers.
DB: "Bit out of my focus area to be honest, but from the past I seem to remember the likely options are direct A-B VoIP peering connections (common between large operators I seem to recall), a central clearinghouse approach like the GSMA's IPX, a third-party one like XConnect, or probably something Internet-based using ENUM, which I've never fully got my head around."
ReplyDeleteBeing brutally frank, I have to say that your response displays a considerable ignorance of the technologies involved and it is hard to take your IMS stance seriously. I think you need to look in detail again at why operators everywhere are deploying IMS and how they think the likely evolution to IP will play out. The reality is that IMS remains the only viable technology choice for large operators and your naive suggestions for IP architecture just don't wash.
Anonymous: you can be as brutally-frank as you like. I don't profess to know everything about everything, but I reckon I have the broadest coverage of any analyst, from spectrum to silicon to enterprise to consumer behaviour.
ReplyDeleteIP peering and VoIP interconnect is not something I have focused on to a great degree, not least because it has historically been irrelevant in mobile.
However, one thing I can be confident about is that almost every network-side IMS specialist I speak to has almost zero knowledge of handsets, and limited knowledge of applications and real-world consumer behaviour.
I have literally heard people say "So, what *is* an IMS-capable handset?", or (amazingly) still use archaic terms like "terminal" as if they lived in the 1980s.
DB:"you can be as brutally-frank as you like. I don't profess to know everything about everything, but I reckon I have the broadest coverage of any analyst, from spectrum to silicon to enterprise to consumer behaviour."
ReplyDeleteBroad your coverage may be but in this instance it lacks the necessary depth in my opinion. It is not enough to simply have a contrarian opinion.
Your comments about having met "network-side"(what exactly that means I'm not sure) people who have little idea about handsets is entirely irrelevant, I'm sure we have all come across similar ignorance at one time or another.
You make some interesting points but I believe that your wholesale dismissal of IMS is based on a misunderstanding of the technology.
I apologise for condemning your opinion anonymously, you provide some very interesting commentary in general but this is one area in which I think you are misinformed.
Anonymous: You are committing a logical fallacy. It is not necessary to understand all parts of a system, in order to say that the system cannot work, or to compare it with other types of system.
ReplyDeleteAn analogy:
You are working as an advisor for hospital administrators, trying to predict future trends in patient admissions and care technology.
On a tour of a medical equipment vendor's facility, a doctor shows you a new MRI scanner, based on the latest technology. It's taken them 10 years to design it.
You pronounce it elegant, but utterly useless. They won't sell any, you say.
In a rage, the doctor accuses you of ignorance and disputes the validity of your opinion, because you admit don't have a detailed understanding of nuclear spin eigenstates.
However, you are still confident of your opinion, as you have noted that the machine has a burnt, frayed electrical plug. And the aperture is too small to fit an average human inside.
He explains that the MRI scanner *must* work, as there's no alternative. Sure, there's the CT machines, but although they're much cheaper they're not optimised as it doesn't provide proper 3D images.
But you know stick to your guns. You know that CT technology is improving rapidly. You've seen new multi-detector machines that can build up multiple images to form a full view. And all the ones you've seen so far are able to fit a person inside. They're not ideal, because of the ionizing radiation - but they do a reasonable job, even though "quality of service" is not cast-iron.
You don't need a PhD in nuclear physics, or an engineering understanding of creating fierce magnetic fields, in order to feel confident. You write a report to your client, the hospital administrators, telling them not to expect many MRI users, and to invest in CT staff instead.
Given the design cycle of the machines is lengthy, you're also confident the situation won't change - especially given what CT machines ought to be able to do, by the time the MRI machine is fixed.
My dismissal of mobile IMS is based primarily on its deficiencies at a service and user-facing level. I published a detailed report on the lack of thought given to IMS-capable handsets over three years ago. I pointed out that devices with "naked SIP" would be available in far greater numbers. Interestingly, even those have been under-utilised, as consumers have been more interested in using mobile browsers and standalone downloadable apps like most on the iPhone. At the same time, they seem quite happy to continue using basic circuit telephony and SMS.
The point of this post was to ask if there's any point in having a cast-iron control plane for services nobody wants.
If I asked you to design an architecture that *just* replaced current CS services, a bare-bones VoIP and messaging platform, do you think we'd end up with IMS? Is IMS really the lowest common denominator for mobile VoIP, as a direct CS replacement? Is there nothing simpler that could be invented, even with scenarios like network-network interoperability?
If you took the worthless "multimedia" M out of IMS, would it still look and cost the same?
Or is it better to refine and keep the CS services, maybe even refresh the equipment? And ditch the near-term vision of all-IP, at least until you can redesign the MRI scanner so it's fit for purpose?
Dean, so lets leave it that you do not like the name and thats all..
ReplyDeletei think IMS is a good acronym anyway..
I think there will have to be an integration of IMS with Internet services because internet applications are more appealing that operators one (i guess thats almost a fact) and operator control on telephony service is somehow desirable (put the case of emergency calls)
But comparing IMS with authoritarian state is simply not a good comparison because authoritarian state is something none wants; but probably libertarian state with the market will regulate himself as major law is starting to be difficult to believe in with the current situation we live.
"If I asked you to design an architecture that *just* replaced current CS services, a bare-bones VoIP and messaging platform, do you think we'd end up with IMS? Is IMS really the lowest common denominator for mobile VoIP, as a direct CS replacement? Is there nothing simpler that could be invented, even with scenarios like network-network interoperability? "
ReplyDeleteYou (and others) seem to have a fixation on the complexity of IMS. The IMS architecture is actually pretty straightforward - a user Database like the HLR (HSS), some call switching functions like MSCs (CSCFs) and some interfaces to application servers (ISC, Diameter etc). Actual implementations may be more or less complex depending on the number of users that need to be supported and network-network interconnects etc.
So the answer to your question above is "yes," if you decided to design a new VoIP architecture for replacement of the PLMN it probably wouldn't look all that different to IMS. This is why just about all new fixed VoIP deployments are now IMS based - it is a straightforward (and standardised) implementation.
> The IMS architecture is actually pretty straightforward - a user Database like the HLR (HSS), some call switching functions like MSCs (CSCFs) and some interfaces to application servers (ISC, Diameter etc).
ReplyDeleteThat's all? Haven't you conveniently forgotten about the other element of the IMS Architecture, the UE (phone)?
Let me get this straight. If the rest of it really *is* as simple as you assert, then I've been absolutely right about the handsets being *the* critical determining factor.
You have inadvertently demonstrated something I have observed many times over the past 5 years - that IMS specialists simply assume that the phone/UE just "works", as if it were just a black-box "terminal" rather than arguably the most important & complex part of the whole end-to-end system.
I used to have a slide which showed the IMS architecture, with all the elements, and the UE tucked away down in the bottom-left hand corner like an after-thought.
Do you reckon Apple's internal app/services architecture diagram looks like that, with the phone relegated to a footnote?
"You have inadvertently demonstrated something I have observed many times over the past 5 years - that IMS specialists simply assume that the phone/UE just "works", as if it were just a black-box "terminal" rather than arguably the most important & complex part of the whole end-to-end system."
ReplyDeleteBah, this is obviously what the whole RCS initiative is trying to address (in a somewhat ham-fisted manner I grant you but it's hardly an assumption that the handset "just works" as you suggest). All operators deploying IMS realise the hurdles that need to be overcome in handset UIs before IMS can work in the mobile environment. All IMS specialists that I have met acknowledge this reality, perhaps you are hanging around the wrong specialists?
All of this does not detract from the fact that IMS is a reasonable VoIP architecture - nowhere near as complex as you think it is and hence popular amongst operators who understand its potential.
If so, how will a user setup a call from his Vodafone end-point to his friend's Orange end-point (could be WiMAX, LTE, Fixed, CS, Google, MSN etc.)?
ReplyDeleteRFC3261 answers this in some detail. Look up the SRV record for orange.com, connect to the IP address given, send 301-INVITE to user@orange, receive 302-REINVITE, initiate RTSP to the endpoint address.
Alternatively, receive target e164 no, reverse and append e164.arpa, look up NAPTR record, and proceed as before.
Or am I missing something?
Being brutally frank, I have to say that your response displays a considerable ignorance of the technologies involved and it is hard to take your IMS stance seriously
...arguments from authority are difficult when you're anonymous.
Hi,
ReplyDeleteIMS also uses the same concept, the only things is that IMS defines a framework including charging, security, authetication etc so that operators can follow and do not have to reinvent it when deploying an all IP based network. Also it assures interoperability of services across n/w borders.
IMS is based on operator centric values delivering on end user expectations and user behaviour.
There are number of ways of doing what u could do with IMS, however to say which approach is better or worse is difficult. End of the day i believe there would be some operators who choose IMS and rest other approach and finally who wins is anybody's bet.
One thing is clear still,at least in our lifetime, most of the operators would continue to depend on standardisation and would choose a standard solution over non standard.....
"...arguments from authority are difficult when you're anonymous."
ReplyDeleteI agree, my comment was rude, I withdraw it. I do think that DB's commentary on IMS is flawed however and his anti-IMS stance is certainly not shared by the parts of the telco industry where I work. My opinion is that the debate over whether IMS will see sufficiently widespread adoption or not is over (for 2 years or more I'd say) and the industry has made its decision despite the nay-sayers.