Monday, December 21, 2009

Will LTE really be an evolution?

A thought struck me on the weekend about the future of wireless networks.

The thought was triggered by a discussion about the release of the Kindle in Australia (and other non US jurisdictions). The Kindle model is an interesting one in which the communications element is built into the device and the subscription and data charges are all paid by the content deliverer (Amazon). In the US they distribut the content over the Sprint network.

To go to the rest of the world they have not entered into separate arrangements with other mobile operators, just relied upon global roaming arrangements for 3G data. This is the explanation for the slightly higher charge for content delivered outside the US.

This started a conversation about the high data roaming rates which are even worse than the high voice roaming rates on mobiles. Part of the justification for these high rates is the strange paths that roaming traffic may need to pursue - especially in the case of voice on the inbound component.

As far as I understand it the signalling layer of mobile networks still sits within the mobile network, so the intelligence to work out where to route the call/data to reach a handset resides in the home network. There is no technical reason why this signalling layer couldn't be exposed to the community of inter-connected operators.

So to use the simple voice call case even domestically, if a customer on the Telstra fixed network is calling a mobile on the Optus mobile network the call is taken by Telstra to the nearest mainland state capital to the calling party and handed over to Optus. The Optus network then determines where on its network (or in the world) the handsert currently is and routes the call. So if the calling party is in Perth and the handset is in Sydney then the mobile operator pays for the Perth to Sydney transport, not the fixed operator. This is the reverse of what occurs in fixed-to-fixed and in part feeds into the higher F2M prices.

Internet routing is different again. As this is a packet network no "connection" is established and each packet can take a different route. The interconnectio is in principle what is known as "hot potato" routing and each network hands packets over to the other network at the first opportunity. However, the question in relation to data on a 3G network is where is the "internet boundary" of the network determined. It is entirely possible that the process is that your data roaming in is all handed back to your home network operator before popping out into the internet.

If someone knows the 3GPP standards (Ian?) and can explain that here that would be good.

Of course, the inefficient routing structure isn't the only reason for the high charges. There are two other factors at least. The first is that high roaming charges are seen to be a very efficient piece of price discrimination - the person who travels is the person who will value the utility of their mobile more and be prepared to pay more overall. The second is just marketing department's preying on customer confusion - the customer when making an initial purchase decision will be driven by the ongoing charges of domestic use and the prices for international use not directly considered, even if over the lifetime of the contract the international roaming charges could exceed the value of the rest.

LTE - which stands for Long Term Evolution - is the standards process being pursuded by the GSM-3GPP tradition. Just as the GSM to UMTS migration jetissoned the TDMA air interface for CDMA, but retained the GSM network control standards, the UMTS to LTE migration jettisons the CDMA air interface for the OFDM interface (as used by WiMAX), but retains the same network control standards.

It is the latter element that makes LTE an attractive migration for 3G operators. The question is will LTE also evolve the network operation standards to make international roaming more efficient?

1 comment:

ian said...

Call routing has not changed from GSM for 3G (AFAIK - I no longer have access to the standards).

There are 2 technical issues for call routing.

1. Calls are routed to the network "owning" the number. Even if the originating network has a mobile signalling stack there is no way to determine if the target is roaming.

Indeed mobile carriers are (rightly) paranoid about signalling, and the cost of interconnecting and testing would exceed any savings which could be made (see below). For a national network the savings would be minimal.

For international roaming the call is routed to the home network, then back to the roaming network. In many cases this incurs 2 international legs for a local call. I still feel that for the low percentage of calls the cost (to the operator) would exceed the savings. There are operators (especailly US) who specialise in offering such services.

2. Once the call hits the network "owning" the number, it is routed efficiently, except in the case the call is not answered when it is routed back to the home network. This case is technically soluble, and CDMA did in fact do this. I believe 3GPP had a proposal to do this, and it may even be implemented in the EU.

Then there are the cost issues.

3. To a carrier with fibre the incremental cost of a long transmission leg are very low, and it makes economic sense to force all traffic onto a fat pipe. Few will realise that most calls on Australian networks route via a switching centre in one of the 6 state capitals or Canberra. A PSTN call from Bathurst to Orange in more likely to go via Sydney than direct. Mobile towers in these locations would be linked back to Sydney, and the call switched on exchanges in Sydney.

4. Establishing an international signalling network for roaming is quite expensive, both in CAPEX and OPEX and utilisation is low. If all national operators would share a single network there could be savings, but the bi-lateral agreements and testing between home and roaming networks remains.

5. The settlement and accounting process contributes to a significant (major) part of the cost of international roaming, but it seems difficult to justify the charges.

Data Roaming is a different issue. (and one I have little experience with)

There are not the routing issues of call roaming. The roaming terminal gets its internet connection from the roaming network, and in principle should (could) be treated similar to a native user, with a premium for higher setup costs. Unfortunately is not quite so simple for those many networks which use real-time (IN) charging.

Again, it is difficult to justify the costs actually charged. Even in a national network operators charge many times the cost per kB that they charge for voice.

My solution to the roaming issue is to purchase a pre-paid SIM for whichever country I am visiting.
This doesn't solve the Kindle issue. It would be more logical for Amazon to establish a commercial agreement with an operator in each country.
This would be subject to business case analysis, but would still cost more than the native US service.