Femtocells Will Rule The Wireless World

Every now and then I think how the future of wireless data will look like. Recent weeks somehow have brought a significant clarification, at least I can see things clearly now... :)

Let's start from the definition of the problem we have. It is simple. Wireless data transmission is too slow. Yes I mean the home and office WiFi networks. Even with the latest 802.11n standard just ratified a few weeks ago, a simple four twisted pair copper CAT5 cable is almost an order of magnitude faster (1 Gigabit per second versus 100-something Megabits) and a couple of times cheaper (a home gigabit switch is $30 today...). But more importantly I mean mobile cellular data. Yes we have 3G and even 3.5G HSDPA networks in place. Delivering on average 1 Megabit per second to mobile terminals. 1 Megabit used to be a high number a while ago. Especially in mobile scenario. But it no longer is considered high today and won't be in the future either. You may say new standards like HSPA+ and especially LTE come to the rescue... Well... LTE rollout will take a lot of time and effort and by the time it is widely available our appetite for mobile bandwidth will be even higher....

So what is next? After LTE ...? Actually not that much... LTE comes pretty close to the limits defined by the most fundamental law of telecommunications - the Shannon Law:

Shannon defines the capacity of a communications channel (C), that depends on the bandwidth of the channel (B) and total received signal (S) versus total received noise (N). The problem we have is B is very limited - it is just a narrow portion of radio waves we can use. And improving on S/N gives us only marginal improvement of C - we have a logarithmic function there. And at the moment we have no idea how, if ever, the C, as defined by Shannon, can be exceeded. So the way to go is to maximize on B, the bandwidth of a channel. Unfortunately B is the most scarce and expensive resource we have.

We cannot afford individual wide transmission channels. There simply is not enough spectrum. That is why channels have to be shared among users. Very early in the morning, say 4 AM, when I log on to a mobile network with my HSDPA modem, very likely I am the only active user. The entire bandwidth of the channel belongs to me, and transmission speed maxes out. But as the day dawns and more users wake up and fire up their modems, they log on to the same base station and we share the channel. With five users logged on, my personal bandwidth is just 20% of the entire channel. We cannot afford individual wide transmission channels.... Or can we?

There is a way to make a wireless channel a private one, that is almost never shared. How? By shrinking the cells. In the beginning we were at 450MHz (NMT) and sCells were big. Some initially had more than fifty miles radius. The capacity of the system was very low. Then came the GSM - initially 900MHz and then 1800MHz and cells were shrinking in order to improve capacity and acommodate more users in densely populated areas. Then came UMTS - 3G with all different transmission scheme - CDMA instead of TDMA. CDMA allows many users share the same frequency and time, so a transmission channel can be wider. Wider channel brings more capacity. More speed. 3G cells are much smaller. LTE - 4G will be even smaller, but they will still be considered large. Large enough to force many users to share them. So with 4G and later 5G an beyond, cells will be smaller and smaller, so small at some point they will be considered almost personal, in the sense there will be just one active user per cell. Or just a few, not a few hundred. We already have small cells available today - the femtocells. But they are considered auxiliary today. To cover dead spots not covered by the large cells. As we move forward, the role of femtocells will be continually increasing, and at some point in time they will heavily outnumber macro cells, actually not even by numbers (that is obvious) but by the area served and by the subscribers served. Millions, or even billions of them.

With such a large number of femtocells, it will be very difficult to manage the intelligence of the network. The physical network architecture will migrate from the intelligent network we have now to the dumb, transparent network, George Gilder has been telling us for years. The intelligence will move to the end terminal, likely based on IPv6 and completely autonomous, not relying on any supporting infrastructure like NATs or firewalls.

Autonomous operation is the key to free roaming and fast switching between the cells. The physical network will be uniform and flat, not hierarchical, like today. Flat, IPv6 address space. And billions of free roaming IPv6 terminals, each with a fully qualified public address. And each secure enough to withstand full public network exposure, as there will be no firewalls , because there will be no single place in the network to place the firewall - it will be flat, not hierarchical like today. Each application on each device will know the trusted peers it should talk to and will drop all other packets. Firewalls will essentially migrate to the end devices.

I have to say I like this vision, as it is simple, consistent and natural. The question, as always, is "when?". Honestly - I don't know, but taking into account the overall acceleration of progress we face today... It may happen before the end of the next decade... Stars are aligning... :)