Monday, 20 September 2010

Logica report on switching costs for M2M (by me)

This last year I've worked on one of the most interesting subjects I've ever come across: Switching costs for Machine to Machine users (or embedded wireless as the GSMA calls it). The result is a report for the Dutch Ministry of Economic Affairs. It's about how a simple rule change in who can apply for numbers (E.212 and telephone) could open up a market for M2M communications. It would more or less remove the difference between public and private networks and result in a more flexible market, like we now already see for internet connectivity.

 Though I independently came to a solution, I can't claim I've thought of this solution first. That honor goes to at least 2 of my Logica colleagues in the Nordics and the UK or maybe to someone at Stratix consulting, but more likely someone somewhere, elsewhere. I do think I can now claim to have written the first full research in a public document on the subject. It's called "Onderzoek flexibel gebruik MNC's"(downloadable pdf, in Dutch). (and yes please put the links that prove this wrong in the comments, someone must have done this before, but I couldn't find it) :-)

It started with a simple question of a customer, but the answer fundamentally questions the business model and regulation of mobile telecommunications. Fortunately the Dutch Ministry of Economic Affairs commissioned Logica (my colleague Jan Lindoff and me) with a report to research this question.

The Question:
How to migrate 10,000 M2M devices from one mobile operator to another mobile operator?

Sounds simple doesn't it? Contract a different operator, receive SIM-cards, switch SIM-cards and presto. Except that the 10k SIM-cards are all over the country in hard to reach places. The logistical nightmare pushed costs up to high heavens. So the customer wanted to be able to do it (in the future) without changing SIM-cards.

For whom?
It's not just one customer, it's everybody. Beer kegs, ereaders, eCall, smart meters, personal navigation devices, cars, photocopiers, containers, trains, fire alarm systems all have been equipped with embedded GSM/GPRS/UMTS. Those not in this list are contemplating whether to embed some wireless.. estimates go into the billions of devices in the next 10 years.

The Answer
With some colleagues I went through the possible solutions until and found that each had a problem associated with it. 
  1. The simple answer seems to be to use the same SIM-card and just switch operators withouth changing SIM-cards. Let the operators fix something in their systems, something like number portability. Unfortunately this doesn't work, because of the way SIM-cards work. Embedded in an unchangeable way on the SIM-card are an IMSI-number (E.212) and several operator specific cryptographic parameters. The first 5 or 6 digits of the IMSI-number are operator specific. These numbers uniquely identify the operator and through this the associated pieces of kit, like a Home Location Register that authenticates the SIM-cards as belonging to a contract having, subscription paying customer. So changing operators would require all mobile operators in the world to recognize that the specific IMSI numbers had changed mobile operator and for these specific numbers to be routed to a different destination. That's already impossible on a global level, but the worst bit is the cryptography. It would require operators to share their cryptographic keys and parameters with competitors. That's a no no. This kills the simple answer
  2. The technical answer is slightly more feasible. What if you could remotely change the details on the SIM-card. Over the air an update is send to the SIM and automagically the IMSI is changed and the security parameters are overwritten with new ones and the device moves to a different network. This is possible. Several companies have proprietary solutions. However, there is no standardized solution and there isn't any mobile operator that supports it. That's problematic as it's the mobile network that determines what SIM's can be used and not the M2M user who wants to switch operator. The 3GPP acknowledged the problem and has worked on functional descriptions of solutions, but until now none of these have found the support of mobile operators (appendix A). Also know that this solution would have to work globally as customers may want to switch all (or some) devices from a European carrier to an Asian one or vice versa. 
  3. The regulatory answer: What if a large scale M2M deployment could use its own SIM-cards; SIMs that carried its IMSI and cryptographic parameters. This way it might not have to change the data on the SIM-card. This is the solution used by true MVNOs today (like Tele2 NL). It would mean it would have to find someone (operator or third party) to host an HLR for it, but this way it could change operators by changing the routing and switching of data instead of going through a logistical or technical process. For this to work the M2M end user should get access to an IMSI-number range, which have to be obtained from a national regulator. The national regulator can assign, within the 3 digit Mobile Country Code of that country (MCC 204 for NL), 2 or 3 digit Mobile Network Codes (MNC) to individual operators or maybe to individual M2M end-users as well
The great thing of the last solution is that it fits in the way GSM was designed it doesn't require any fundamental technical changes. It might however require changes in political and business thinking. Not only does this last solution allow for 'easy' switching, it also seems to solve some other problems large scale M2M users have. For instance:
  • Coverage issues: This is a big problem. Almost by definition there is perfect coverage everywhere, except in the 4 square meters where the device is located. Worse, the other networks have perfect coverage except the one chosen for this deployment. This is an unsolvable problem, because wireless propagation is voodoo. Only in the field you know what coverage you have at that moment, a new building, a change in the orientation of an antenna etc. can change it in the future. By contracting (national) roaming on multiple networks in a geographic area an M2M deployment could increase its coverage. (This is the same as when you go abroad and generally can use all available networks instead of just the one you have your contract with. A Vodafone NL customer can roam on Orange UK, but a Vodafone UK customer cannot. Some M2M users actually use international SIMs for this reason)
  • International roaming: Those M2M users whose devices travel: personal navigation, eReaders, cars, could contract different operators in different geographic regions. (At this moment thought this might be hard to arrange as roaming agreements seem to be more or less closed to non-GSM network owners)
  • Initial provisioning/Lifecycle management: Large scale M2M deployments often want to use one device on a global scale. You could think of an eReader with an embedded communications device or a car. For communications cost reasons this may not be smart. So they are now torn between producing just one device in China and distributing this globally or producing country/region specific devices. Many retailers however don't like country specific devices, they want to distribute inventory according to need and not according to how the supplier pre-packaged it. Furthermore those nasty customers may buy a German version and use it in The Netherlands, which either means crippling the functionality while the user is abroad or accepting huge roaming fees. 
So the last answer looked very promising, but the big question was: Why isn't anyone doing this? It looks so easy. I'm old enough to know that other people are more experienced than I, so if no one is doing it, there must be a good reason for it ... and there is. There is a snag in the regulations regarding E.212 numbers (as IMSI's are also known). The Dutch (and as far as I know all European) regulations require the user of such a number to be a "public" provider of electronic communications networks, which many M2M end users can't claim to be.

Being an ex-regulator I know regulations can be changed, so I talked to the government, who happened to have been informed of the solution by a side note in a report of Stratix. In the end we won a commission in competition to evaluate the solution in full. What was really nice about this was that it touches on some core issues in mobile telecommunications:
  1. Why is it that end-users don't have access to wholesale markets for (mobile) telecommunication. Almost every market allows significantly large users to access the wholesale market. In energy every farmer can become an electricity or gas producer, they can access the spot trade market etc. The same goes for banking services where lease companies run banks and trade on the international markets. In telecoms this is only really possible for internet peering and transit. The fixed telephony market is semi-locked and the mobile telephony market is fully locked for large scale end users. 
  2. How come roaming is expensive? If all that distinguishes one mobile device from another is the IMSI number used, than why is it expensive to use an Orange UK SIM on an Orange France network? It could be the same HLR and network serving the customer. If foreign just means that the first three digits of the SIM are different, it's not much of a distinction. So why is there a distinction?]
  3. What does it mean to be a public network and why do we have the distinction in law between public and private networks? If it is such an important difference, then why do private networks like Google, BBC and Microsoft have no problems connecting to the internet. I used to love this debate, when I was discussing lawful interception and data retention, but this is another view on the same problem. 
  4. How to use Private GSM in the DECT Guard Band, which is possible in the UK and NL already. Private GSM makes use of some spectrum that was reserved as a guard band between DECT and GSM 1800. In The Netherlands the use is license free. It allows users to make use of normal GSM's to connect to a private low power network which could replace the use of DECT, which often has horrible propagation characteristics. A problem is getting access to SIM-cards for this use. 
I won't comment on the specific results of the report, it should speak for itself.

All in all, I've had tremendous fun researching these questions. Please read the report, all 55 pages or the slides, if you can't read Dutch. If you have any questions post them in the comments.


  1. The main issue I see with solution 1 is the limited number of MNCs. I doubt that country codes will be combined with MNCs, there's just too much at stake at the ITU level, hence only 3 digits per country will be available for M2M vendors. Hence, MNCs will become rare and some intermediaries are likely to trust MNCs, which recreates the issue of MVNOs. Couldn't one alternative be to piggyback on number portability regulations— of course changing the way number portability is provided ?

  2. @jr unfortunately I hear this misunderstanding often. There really is no shortage of MNC's. End-users need only one for it to work globally. So there is no reason to use national MNC's of each country. Even if just the EU would combine their MNC's there would be roughly 20,000 3 digit MCC+MNC combinations possible. Or if the ITU would decide to utilize the 900 range for this use, there would be close to 100,000 MNC's available. Furthermore experience from the internet, where companies can get AS numbers already, shows not everyone wants to do this. All in all there is no reason to expect a shortage in a very long time and we would have ample opportunity to deal with a shortage once we run out of the first 20k, or 100k MNC's.

    I don't know what you mean with the issue of MVNO's, so I park that one.

    Piggybacking number portability isn't really practical here. First of all, only the first 6 digits are looked at for routing decisions. What you propose would mean routing on individual numbers. Furthermore, number portability regulations are national. The use of IMSI's is global. In case a french E.164 number changes operator eg. Neuf to foreign operators still know it's French by the first digits and know where to dump their traffic. In case of an IMSI that has changed operator this isn't possible. This would require every operator in the world to participate, otherwise things like roaming don't work.

    The fundamental difference is that an E.164 number just routes traffic at layer 7. An IMSI authenticates devices and traffic at layer 2.

  3. Indeed combining MNCs EU wide can help, but 20,000 codes are still limited compared to the number of companies in the EU; all I'm saying is that this solution is IPv4-like in a world that needs IPv6. Given IPv4's success, this must be read as a compliment ! And you can use tricks derived from IP such as NAT in order to go around short term limitations.

    The MVNO issue I refer to is the fact that M2M device manufacturers do not want to be held hostage by network operators; if these operators take or have control over scarce resources such as MNC, they can put a hefty price tag on these. In fact, my point is that the issue appears as soon as anyone can trust MNC codes.

    Finally, on number portability (E.164 based), I stated that number portability principles would have to be changed which is what you point at in your response. Indeed E.164 is to IMSI what an IP address is to a MAC address (sorry to oversimplify - I know IP is way below layer 7). The success of IP addressing should make us think twice before going for mechanism akin to MAC addressing.