.comment-link {margin-left:.6em;}

Monday, October 30, 2006

IMS is all back to front

In August the IEC, which runs major events for the telecoms industry among other things, published an article entitled "IMS – Putting Value Back into the Network and Increasing the Revenues Flowing Out" by Telcordia, an IMS vendor. Full transcript here.

It concluded: "Now that the long awaited 'network of networks' looks like it's finally emerging from the complex cat's cradle of coexisting and often competing technologies and protocols that have grown up in recent decades, it's important to remember that IMS is there as a true business enabler. Just as the invention of money revolutionized entire economies and social structures, replacing the inevitable time and space limitations imposed by bartering, so too is IMS set to open up the communications environment to new ways of doing business. In the process, new value–and new wealth–will be created."

We were stunned that this level of vendor hype still exists! Our large survey earlier in the year showed that people in the know really don't believe this. And our industry brainstorm in October - where for the first time at a major event all the participants were allowed to debate the issue anonymously and simultaneously in a structure manner and in real-time - confirmed very clearly that
The IMS Emperor (as originally promoted by vendors and, it seems, still promoted by some) Has No Clothes. It's amazing what can happen when you let everyone to have their say without fear of ridicule... Summary: There's plenty of technical usefulness within IMS, but not to the degree espoused here.

I asked our Chief Analyst, Martin Geddes, for his view:

Essentially, we at STL and Telco 2.0 disagree with Telcordia's premise. Yes, there are APIs needed into authentication and other services. But a lot of this stuff doesn't need the IMS architecture. You just do it over the current Internet, and roll a suitably lightweight protocol to exchange the data.

Things like QoS have been tried and largely rejected before. The problem is that asking all the nodes in the network to reserve some capacity takes time. Indeed, so much set-up time, that it becomes a problem. If a telco has one advantage over the Internet, it's the ability to deploy boxes geographically/topologically scattered around the network, to minimise latency. This is how cellular works: many HLRs and switches, rather than backhaul to a central location. We've already been here before with ATM (an earlier telecom protocol that many wrongly thought would displace IP.)

We've thrashed over the other supposed "benefits" of IMS before (see previous postings in this blog) -- the words sound good, but the business case is hollow because the same end-user value can be delivered too consistently over the Internet.

Telcordia in their article say:
"The prime purpose behind IMS is to enable the seamless convergence of all the communications services that we currently use but are partitioned by the nature of the networks that they run on. While we have become used to using fixed Internet for some transactions, our mobiles for others, and so on, this silo concept is increasingly inefficient and expensive for both user and service provider. ...

With traditional voice revenues under constant erosion, it is essential that service providers of all types are able to move up the value chain, away from basic connectivity and toward more advanced communications services that include multimedia, messaging, and business and lifestyle applications."

But Telcos have no history of success in rich application development. Furthermore, the assumption that you move *up* the value chain is Telco 1.0 thinking. The world is flattening, and most operators are going to have to relinquish dreams of being more than a utility pipe. The rich multimedia stuff is also contextual (e.g. as part of a game), and the owners of those services have no intention of sharing the pie with the telcos when there's an alternative, cheaper distribution mechanism (the Internet).

There are two possible exception areas: We basically agree with Telcordia's security pitch, although again I'm not convinced IMS is the right architecture. And the PSTN/legacy interconnect issue won't go away.

Old, dumb, edge nodes aren't going to suddenly get smart, so you need some smart network node to act on their behalf to integrate them into new services. But my desk phone isn't about to spout a screen, so
forget those multimedia combinational services. They'll emerge from the PC world (and indeed already have).

Fine-grained billing isn't loved by users. Predictability is in, and flat rate is one way to deliver it. The dream of a telco is to have a zillion rating rules, and for it to be impossible for users to comparison shop.

Ultimately, IMS is all back-to-front. Rather than trawling all the network traffic trying to find billable events, you just offer a bunch of service APIs that can be invoked under the control of the edge points (or third party servers). The billable event, if any, is under the control of the third party. Session control in the network doesn't add much value, and ultimately can't be charged for. Skype proves it -- we talked in wideband audio yesterday, no QoS or IMS needed.

Here's the crux: the telecom industry believes there are (i) new, mass-market services along the lines of voice telephony and SMS to be built, and (ii) they will be sold along the same lines -- metered service, connectivity puchased as part of the services and not separately provisioned (as with broadband).

We think #1 is false. There are some extensions to existing talk and messaging apps that will "go universal". But not many. There may be a few IMS-enabled apps that become popular within single carriers or regions (PoC, Video Sharing), but it'll be an exception.

There never will be a global, interoperable push-to-talk network from the telcos.
There will never be a telco IM system independent of the existing Internet portal players. There isn't a "one size fits all" conferencing system out there waiting to be built.

However, #2 is probably something that we'd want to move forwards with still. The end-to-end principle that defines Internet architecture says nothing about how the pipe is sold.

So there's probably a simpler architecture than IMS still waiting to be built. Telcos would still be involved in accounting for network traffic and billing it to either the user or application providers (or advertisers). But they wouldn't know anything about what the traffic really represents.

IMS can be used to do this, but it's a horribly complex way of doing it.

An alternative approach could be "a billion Internets". Create an endless succession of copies of the Internet address space. It's still a dumb (virtual) network, but you've got some kind of control over who is admitted. "Internet #18237232" might be for a particular purpose
like gaming. It gets a certain priority of service, and network traffic is billed to particular nodes. It's a bit like Paris Metro Pricing (see the IMS Insider Summer Report), but on a grander scale -- or on existing enterprise VLAN (virtual LAN) technology scaled up globally.

I think we'll do an article on this for our next IMS/Telco 2.0 Quarterly Report...

(In the meantime, a far more pragmatic discussion here on Telecom TV by some friends of our's).

The Editor, IMS Insider

The 'billion internets' thing is pretty close to the 802.1ah standard (PBB). Not sure if it envisages a billion of them, but it certainly sounds like the concept you are proposing.
Post a Comment

<< Home

This page is powered by Blogger. Isn't yours?