Monday, December 24, 2001

Building a Wireless Internet Access Infrastructure

How will the wireless Internet be built? Will it evolve from today's digital wireless cellular phone networks, i.e. GSM, GPRS and finally UMTS? Or will it be WLANs? Will it be a few nationwide network infrastructures built by large carriers (today's telecoms), or will it be an interconnected network of islands owned by any size and kind of network providers similar to the structure of the Internet itself?

The nationwide-carrier-model seems to be limited by physical restrictions. The model does work economically only if ubiquity can be achieved by a limited number of base stations. This did work with GSM. A single base station can cover an area of several kilometers in diameter. With UMTS it does already get difficult. A UMTS macro cell, which is several kilometers in diameter has a maximum transmission rate of only 144 kBit/s. This is not so much more than GPRS and also degrades if there are many users inside the same cell. UMTS micro cells (<1km) support up to 384 kBit/s and only the pico cells (<30m) support 2MBit/s. However, if the pico cell is shared by 9 users, each user has only 64kBit/s of bandwidth. If you move faster than 10km/h you are referred from the pico to the micro cell, if you move faster than 120km/h to the macro cell. Considering this, the advantages of UMTS over GSM+GPRS seem small.

WLANs offer 2 MBit/s or more and have a similar reach like UMTS's pico cells. The problem is that both the UMTS pico cells and WLANs have a reach that is too small to allow carriers to economically build network infrastructures with nation-wide coverage (it would work if only one infrastructure would be built by a monopolist, but that's history...).

In summary there is a choice between nation-wide wireless network infrastructures that are limited to bandwidth in the 9.6 to 64kBit/s range on the one hand, and WLAN networks that provide >2 Mbit/s but are limited to "hot spots", i.e. places frequently visited by many people, your office and your home. Actually it's not a choice, the solution is to combine both infrastructures:
  • the national wireless infrastructures provide ubiquity, but a limited bandwidth that is sufficient primarily for asynchronous messaging (email, SMS) and bandwidth- and latency-optimized synchronous applications like chat/Instant Messenging.

  • WLANs provide bandwidth that is necessary for pleasant user experiences with web surfing, video conferencing and mobile VPN access to corporate networks and enterprise applications.

This combined infrastructure has only one problem: without a further infrastructure there is no roaming and no billing between the WLAN islands. I think it is unlikely that this can be solved by direct roaming agreements between WLAN operators. This will not scale because there will be too many of them. It's not just cafe's, hotels and airports. What about schools, universities and hospitals? Neighbourhoods that today share satellite dish or TV cable and tomorrow have a WLAN too? Corporate WLANs? What seems to be required here is an intermediary that is neither a network operator nor a hot-spot / public / private / semi-public space owner. Someone who hides the complexity of the different authentication and billing schemes of each WLAN operator. I think this is what [Boingo] is doing. Here are two interesting articles:

They provide seamless roaming by making billing, authentication and confidentiality/encryption of the transmitted data transparent to the user. I think this is a simple but brilliant idea. They are service providers only, not network providers. Even corporate networks, neighbourhoods etc. could have peering agreements with them. Finally these few national/international service providers will have peering agreements between them.