Sunday, April 30, 2006
IMS Reality Check by Martin Geddes
Martin Geddes has been invited as guest blogger by IMS Insider to give an IMS Reality Check.
Martin is commenting on 7 promises made by IMS (I give here only an abbreviated version of his replies, for the full text please look here)
Promise # 1. New Revenue Opportunities: From silo to a horizontal business model
IMS is partly counter-productive, because tying functionality and value to the network platform via SIP proxies (as opposed to the business platform via web services) preserves the network silo...
Promise # 2. New services: Multimedia rich services that make operator's offerings "stickier"
Media is increasingly based on file transfer, not streaming, so IMS has little (if any) value to add over existing broadband technologies...
Promise 3. Faster time to Market: Standard network elements drastically decrease application development and rollout
A Faustian bargain. By embedding intelligence in the network, you're still not capturing the agility of edge-based innovation ....
IMS isn't exactly going to be the platform of choice for an upcoming generation of developers – too complex, too much effort dealing with carriers and their caprice...
[Richard's note: regarding complexity, see David Meyers piece on Back to the Future. IMS is simply too complex, way off in the long tail.]
Why has Intelligent Network only delivered a handful of features? If you can't do more than caller ID, 3-way calling and call forwarding in 20 years on IN, what is the hope that you'll do better on IP/IMS?
[Richard's note: But IMS is seen by some operators an IN 2.0. statement from a representative from TeliaSonera: We are building with IMS the same environment on IP what is SS7 today! And he meant this seroius.]
Promise 4. Investment Protection: Future service compatibility is assured through established interfaces
Possibly. ....
Promise 5. Lower OPEX: Through infrastructure reuse for new services
In principle, yes. But... you're possibly running a fully-depreciated network with trained staff, and can cannibalise parts and re-negotiate support and maintenance costs with vendors. So there is some degree of IMS alternative. Hey, why not skip a generation and go straight to peer-to-peer or other edge-based intelligent devices! Same outcome, 1/1000th the cost.
[Richard's note: so definitely not, all IMS investments will be stranded cost]
Promise 6. Fixed Mobile Convergence: Similar systems are used for fixed or mobile and can be bridged to offer ubiquitous service
Yes, to the extent mobile continues as a semi-closed ecosystem....
Promise 7. Higher control of the network: Operators become service providers instead of fat pipes.
*Instead* of fat pipes! Weep for their investors... Is IMS really just a desperate attempt to create billable events and scarcity-meter network usage?
[Richard's note: Yes, it is. It is all about billing and charging, as I pointed out in my comment to
"The 3G IP Multimedia Systems (IMS), Second Edition" from Gonzalo Camarillo and Miguel A. Garcia-Martin.
Now here comes the important part:]
The customers are taking control. The only way is down in terms of control. Customers want fat, always-on and open pipes to run the applications of their choice. They also want something that works, is simple to buy, can be supported, and doesn’t require them to perform complex set-up – but that’s not predicated on network control, but rather on brand values, product integration and channel control.
Control of wireline networks is almost totally gone in open, competitive markets in free nations. The feudal overlord model is under attack on wireless too: eroded by new technology (e.g. WiFi, Flarion, IPWireless, WiMax, etc.), new entrants (e.g. recent Intel/Pipex WiMax annoucement yesterday), and competition among existing operators (e.g. the 5 cell networks in the UK). The simplest, cheapest way of differentiating yourself is to become more open (e.g. T-Mobile Web’N’Walk) – a ratchet towards ‘open’.
IMS doesn't achieve this goal unless you have significant market power to force customers through the IMS toll gates. A good model for monopoly markets with licensed carriers, maybe – but that’s hardly the wave of the future. Maybe you can ‘back-fill’ markets like WiFi roaming by offering service-specific access to people who aren’t interested in a full day’s WiFi subscription (e.g. like Skype Zones, where you just get Skype access). Even then, it’s slow and complex.
[Richard's note: the ghost is already out of the bottle]
There is an IMS alternative, but aims to a fundamentally different goal. Deliver super-abundance; where capacity limits exist, innovate in terms of application-agnostic QoS mechanisms, and hunt for better ways of pricing and funding networks that don’t rely on opening up the packets to see what’s written inside.
[Richard from here on
Martin does not even mention three other IMS myths most proponents are counting under the promises (this would have made it 10) : the QoS myth, the reliability myth and the security myth.
There is a follow-up post on IMS Insider by Christophe Gourraud from Swisscom. He is saying:
The fact that IMS is a telco specification does not mean that it defines a new telco world apart from the Internet. It is true that there is an IMS network architecture, and that there exist IMS-specific extensions to SIP. However, this architecture and these extensions do not prevent simple interworking between SIP devices and applications in the IMS domain and SIP devices and applications in the Internet space.
Good point. The real problem I have with IMS is that really nobody could explain to me yet what I can do with IMS what I cannot do with "plain" SIP. Even SIP gets already too complex for my taste.
What I like in the Swisscom reply is:
Likening IMS to IN is a big mistake. IN permits to plug application servers on top of a voice-centric network through a dedicated telco protocol which essentially permits to control calls. On the other hand, the IMS service architecture can be seen in a nutshell as a way to enrich IETF-based SIP routing with (user or service) profile-based mechanisms. Once you understand that SIP is not the typical ISUP or INAP/CAP, and that it can support much more that session control, alone or with sister Internet protocols, you can start to think otherwise about IMS.
Hear, hear, TeliaSonera.
What is IMHO missing in all these statements is the only asset IMS really has, as a takeover from GSM: it is identy = SIM = ISIM.
So if we forget about all these 30-40 functions in IMS and just stick with the S-CSCF, the HSS and the SIM-Card, IMS would be fine with me and bring a benefit to everybody involved, including the end-user.]
Martin is commenting on 7 promises made by IMS (I give here only an abbreviated version of his replies, for the full text please look here)
Promise # 1. New Revenue Opportunities: From silo to a horizontal business model
IMS is partly counter-productive, because tying functionality and value to the network platform via SIP proxies (as opposed to the business platform via web services) preserves the network silo...
Promise # 2. New services: Multimedia rich services that make operator's offerings "stickier"
Media is increasingly based on file transfer, not streaming, so IMS has little (if any) value to add over existing broadband technologies...
Promise 3. Faster time to Market: Standard network elements drastically decrease application development and rollout
A Faustian bargain. By embedding intelligence in the network, you're still not capturing the agility of edge-based innovation ....
IMS isn't exactly going to be the platform of choice for an upcoming generation of developers – too complex, too much effort dealing with carriers and their caprice...
[Richard's note: regarding complexity, see David Meyers piece on Back to the Future. IMS is simply too complex, way off in the long tail.]
Why has Intelligent Network only delivered a handful of features? If you can't do more than caller ID, 3-way calling and call forwarding in 20 years on IN, what is the hope that you'll do better on IP/IMS?
[Richard's note: But IMS is seen by some operators an IN 2.0. statement from a representative from TeliaSonera: We are building with IMS the same environment on IP what is SS7 today! And he meant this seroius.]
Promise 4. Investment Protection: Future service compatibility is assured through established interfaces
Possibly. ....
Promise 5. Lower OPEX: Through infrastructure reuse for new services
In principle, yes. But... you're possibly running a fully-depreciated network with trained staff, and can cannibalise parts and re-negotiate support and maintenance costs with vendors. So there is some degree of IMS alternative. Hey, why not skip a generation and go straight to peer-to-peer or other edge-based intelligent devices! Same outcome, 1/1000th the cost.
[Richard's note: so definitely not, all IMS investments will be stranded cost]
Promise 6. Fixed Mobile Convergence: Similar systems are used for fixed or mobile and can be bridged to offer ubiquitous service
Yes, to the extent mobile continues as a semi-closed ecosystem....
Promise 7. Higher control of the network: Operators become service providers instead of fat pipes.
*Instead* of fat pipes! Weep for their investors... Is IMS really just a desperate attempt to create billable events and scarcity-meter network usage?
[Richard's note: Yes, it is. It is all about billing and charging, as I pointed out in my comment to
"The 3G IP Multimedia Systems (IMS), Second Edition" from Gonzalo Camarillo and Miguel A. Garcia-Martin.
Now here comes the important part:]
The customers are taking control. The only way is down in terms of control. Customers want fat, always-on and open pipes to run the applications of their choice. They also want something that works, is simple to buy, can be supported, and doesn’t require them to perform complex set-up – but that’s not predicated on network control, but rather on brand values, product integration and channel control.
Control of wireline networks is almost totally gone in open, competitive markets in free nations. The feudal overlord model is under attack on wireless too: eroded by new technology (e.g. WiFi, Flarion, IPWireless, WiMax, etc.), new entrants (e.g. recent Intel/Pipex WiMax annoucement yesterday), and competition among existing operators (e.g. the 5 cell networks in the UK). The simplest, cheapest way of differentiating yourself is to become more open (e.g. T-Mobile Web’N’Walk) – a ratchet towards ‘open’.
IMS doesn't achieve this goal unless you have significant market power to force customers through the IMS toll gates. A good model for monopoly markets with licensed carriers, maybe – but that’s hardly the wave of the future. Maybe you can ‘back-fill’ markets like WiFi roaming by offering service-specific access to people who aren’t interested in a full day’s WiFi subscription (e.g. like Skype Zones, where you just get Skype access). Even then, it’s slow and complex.
[Richard's note: the ghost is already out of the bottle]
There is an IMS alternative, but aims to a fundamentally different goal. Deliver super-abundance; where capacity limits exist, innovate in terms of application-agnostic QoS mechanisms, and hunt for better ways of pricing and funding networks that don’t rely on opening up the packets to see what’s written inside.
[Richard from here on
Martin does not even mention three other IMS myths most proponents are counting under the promises (this would have made it 10) : the QoS myth, the reliability myth and the security myth.
There is a follow-up post on IMS Insider by Christophe Gourraud from Swisscom. He is saying:
The fact that IMS is a telco specification does not mean that it defines a new telco world apart from the Internet. It is true that there is an IMS network architecture, and that there exist IMS-specific extensions to SIP. However, this architecture and these extensions do not prevent simple interworking between SIP devices and applications in the IMS domain and SIP devices and applications in the Internet space.
Good point. The real problem I have with IMS is that really nobody could explain to me yet what I can do with IMS what I cannot do with "plain" SIP. Even SIP gets already too complex for my taste.
What I like in the Swisscom reply is:
Likening IMS to IN is a big mistake. IN permits to plug application servers on top of a voice-centric network through a dedicated telco protocol which essentially permits to control calls. On the other hand, the IMS service architecture can be seen in a nutshell as a way to enrich IETF-based SIP routing with (user or service) profile-based mechanisms. Once you understand that SIP is not the typical ISUP or INAP/CAP, and that it can support much more that session control, alone or with sister Internet protocols, you can start to think otherwise about IMS.
Hear, hear, TeliaSonera.
What is IMHO missing in all these statements is the only asset IMS really has, as a takeover from GSM: it is identy = SIM = ISIM.
So if we forget about all these 30-40 functions in IMS and just stick with the S-CSCF, the HSS and the SIM-Card, IMS would be fine with me and bring a benefit to everybody involved, including the end-user.]
Comments:
Richard, I couldn't agree more with your conclusion on what an actual benefit of IMS could be. Identity using ISIM. Next is more culture than technology, so I am skeptically optimistic...
The second benefit I see from IMS is the seperation of rather clear definition of a SIP proxy (S-CSCF) and the HSS as seperate functional elements.
The second benefit I see from IMS is the seperation of rather clear definition of a SIP proxy (S-CSCF) and the HSS as seperate functional elements.
Permalink posted by 
: 04 May, 2006 14:26
: 04 May, 2006 14:26
IMS is for me two things: 1) some parameters and protocol aspects that are not covered by plain SIP, especially in a mobile environment. And 2) - more importantly - a common framework / architecture how to apply all the IETF and telco protocols.
Let me explain why a framework is as important as the protocols you use. For example SPAM. There are at least - let's say - 4 technical approaches how to get rid of SPAM. What is missing is a common framework for all ISPs how the Internet communitiy is going to solve it.
Coming back to IMS. The wonderful "two layer" architectural separation of P-CSCF and S-CSCF
allowing all market players to benfit from IMS: today, without this concept, every ISP rolls out its own VoIP infrastructure (S-CSCF) and connects it to the telephone network via gateways. In the best case, some of these ISPs build up a partnership and allow calls between them without going through gateways. But the "roaming" of VoIP (or generally SIP communication) is not solved. Each VoIP provider will have a contract with each other VoIP provider he allows to make calls to or receive calls from. The number of contracts will explode
By applying the IMS architecture, a standardized "SIP roaming" is possible between the P-CSCFs and the the S-CSCFs. In the future, I foresee a kind of new market player, the missing link, who provides this SIP-roaming network.
Let me explain why a framework is as important as the protocols you use. For example SPAM. There are at least - let's say - 4 technical approaches how to get rid of SPAM. What is missing is a common framework for all ISPs how the Internet communitiy is going to solve it.
Coming back to IMS. The wonderful "two layer" architectural separation of P-CSCF and S-CSCF
allowing all market players to benfit from IMS: today, without this concept, every ISP rolls out its own VoIP infrastructure (S-CSCF) and connects it to the telephone network via gateways. In the best case, some of these ISPs build up a partnership and allow calls between them without going through gateways. But the "roaming" of VoIP (or generally SIP communication) is not solved. Each VoIP provider will have a contract with each other VoIP provider he allows to make calls to or receive calls from. The number of contracts will explode
By applying the IMS architecture, a standardized "SIP roaming" is possible between the P-CSCFs and the the S-CSCFs. In the future, I foresee a kind of new market player, the missing link, who provides this SIP-roaming network.
Permalink posted by : 31 May, 2006 19:46
: 31 May, 2006 19:46
I agree that a common framework (profile) is required for SIP interoperability. IETF SPEERMINT is trying to achieve this with say 20 active people. 3GPP IMS and ETSI TISPAN is trying to achieve the same with about 1000. ETSI TISPAN alone produced 1500 temporary documents last years simply in endorsing 3GPP documents.
It is also unclear to me what the "wonderful" separation of P-CSCF and S-CSCF is. The P-CSCF is basically an outbound proxy also possible in plain SIP. Considering that P-CSCFs in visited networks is planned only in 3GPP release 7 to implemented earliest in two years, I wonder.
And I do not understand the concept of "roaming" on the Internet. I am using my SIP proxy from anywhere in the world, I do not need an "visited" network for this.
As Henry Sinnreich says: "This is all optional"
It is also unclear to me what the "wonderful" separation of P-CSCF and S-CSCF is. The P-CSCF is basically an outbound proxy also possible in plain SIP. Considering that P-CSCFs in visited networks is planned only in 3GPP release 7 to implemented earliest in two years, I wonder.
And I do not understand the concept of "roaming" on the Internet. I am using my SIP proxy from anywhere in the world, I do not need an "visited" network for this.
As Henry Sinnreich says: "This is all optional"
Okay, I agree. The paperwork they produce is rather big and the amount of people involved is definitvely too high.
Questions:
1) I assume your SIP provider bases its service on an existing telephone network as "fallback". If that "legacy network" doesn't exist any longer, what does he do? IMS is not thought as an overlay on an existing infrastructure but building a new era.
2) Can your grandma buy a SIP phone anywhere, plug it in and it works? A global framework would be nice, I mean covering more than SIP. e.g. how to prevent the very boring "SIP spam" that starts to grow
3) If you are in a Hotel in Burundi and invite your SIP to connect to 911, what happens? I see this as one type of the "roaming" issue.
Questions:
1) I assume your SIP provider bases its service on an existing telephone network as "fallback". If that "legacy network" doesn't exist any longer, what does he do? IMS is not thought as an overlay on an existing infrastructure but building a new era.
2) Can your grandma buy a SIP phone anywhere, plug it in and it works? A global framework would be nice, I mean covering more than SIP. e.g. how to prevent the very boring "SIP spam" that starts to grow
3) If you are in a Hotel in Burundi and invite your SIP to connect to 911, what happens? I see this as one type of the "roaming" issue.
Permalink posted by : 31 May, 2006 22:19
: 31 May, 2006 22:19
One remark why I think the "one-proxy-is-enough-approach" current NGN/VoIP deployments use today will most probably not work for a growing customer base:
Just think of the reasons why file server services invented the concept of "mirros". Isn't there an analogy to "visited network" vs. "home network"?
Just think of the reasons why file server services invented the concept of "mirros". Isn't there an analogy to "visited network" vs. "home network"?
Permalink posted by : 31 May, 2006 22:53
: 31 May, 2006 22:53
To your questions:
1) I assume your SIP provider bases its service on an existing telephone network as "fallback". If that "legacy network" doesn't exist any longer, what does he do? IMS is not thought as an overlay on an existing infrastructure but building a new era.
What do you mean with fallback? My SIP provider uses his DSL line as access, my number is ported -so I have an E.164 number, it uses a Sipura TA (but I could use a Softclients as well - I have all the configuation data, you can reach me also via a SIP URI (sip:+4319793321@sip.sil.at), I have an ENUM entry for it in e164.arpa and my provider also queries ENUM on each outgoing call. Of course it needs the PSTN to interconnect with numbers on the PSTN and all VoIP providers not providing their customers with SIP URIs (like everybody else). If you think that the IP Interconnect problem is solved with IMS per se, you should participate in 3GPP and TISPAN to get the message how far they are down this road.
2) Can your grandma buy a SIP phone anywhere, plug it in and it works? A global framework would be nice, I mean covering more than SIP. e.g. how to prevent the very boring "SIP spam" that starts to grow.
In case of my provider it worked out of the box without any configuration. They only item I had to enter myself was the WEP-key for WiFi access.
3) If you are in a Hotel in Burundi and invite your SIP to connect to 911, what happens? I see this as one type of the "roaming" issue.
We are working in IETF ECRIT to solve these issue. It will work with or without IMS. BTW, I will need 112 or 133.
1) I assume your SIP provider bases its service on an existing telephone network as "fallback". If that "legacy network" doesn't exist any longer, what does he do? IMS is not thought as an overlay on an existing infrastructure but building a new era.
What do you mean with fallback? My SIP provider uses his DSL line as access, my number is ported -so I have an E.164 number, it uses a Sipura TA (but I could use a Softclients as well - I have all the configuation data, you can reach me also via a SIP URI (sip:+4319793321@sip.sil.at), I have an ENUM entry for it in e164.arpa and my provider also queries ENUM on each outgoing call. Of course it needs the PSTN to interconnect with numbers on the PSTN and all VoIP providers not providing their customers with SIP URIs (like everybody else). If you think that the IP Interconnect problem is solved with IMS per se, you should participate in 3GPP and TISPAN to get the message how far they are down this road.
2) Can your grandma buy a SIP phone anywhere, plug it in and it works? A global framework would be nice, I mean covering more than SIP. e.g. how to prevent the very boring "SIP spam" that starts to grow.
In case of my provider it worked out of the box without any configuration. They only item I had to enter myself was the WEP-key for WiFi access.
3) If you are in a Hotel in Burundi and invite your SIP to connect to 911, what happens? I see this as one type of the "roaming" issue.
We are working in IETF ECRIT to solve these issue. It will work with or without IMS. BTW, I will need 112 or 133.
Addition: I never got a SIP SPAM message on my phone yet, and it is now one year on-line.
Correction: the SIP URI is sip:+4319793321@enum.sil.at
Correction: the SIP URI is sip:+4319793321@enum.sil.at
Re: one-proxy-is-enough - simple proxies can take quite a lot of subcribers and there are other means to increase the load. Nevertheless, you are correct that the S-CSCF (in combination with the HSS!) is a nice way to implement a large customer base - as Colin also stated in his first post (and also in his presentation at the Spring VON). The HSS (and the ISIM) is also IMHO the only advantage the IMS is providing.
Post a Comment
