Tuesday, May 17, 2011

On ice skating and open bands !


Unlicensed bands are like skating rinks

If you have never seen or heard about a public skating rink, you could think the whole idea was crazy - toddlers, grandparents, teenagers, speed maniacs, all without helmets or knee-pads skating in a common area with no lanes, minimal rules and no guarantees that you won't get hurt! But the fact is, it does work and the reason it works is precisely the absence of specific rules.

Open wireless bands such as the 2.4 GHz ISM bands too have a large common spectrum-area, different kinds of participants - high bandwidth WiFi, narrow band but hopping natured Bluetooth, etc. It has a few transmission rules and most importantly just like the skating rink, no guarantees about interference from other users. You can and probably will collide but overall the system seems to work fairly well.

The idea of spontaneous order

Daniel B. Klein speaks about the same skating rink analogy with respect to society and economy. He argues that "intuition leads us to think that complex problems require complex, deliberate solutions. In a roller rink, the social good depends on getting the patterns to mesh. But no one is minding that good..... but in promoting my interest in avoiding collision with you, I also promote your interest in avoiding collision with me." And that is probably the key: coincidence of interest

Imagine a "rink master" in the rink who sits at the center and communicates instructions to individual skaters - 'Move right in 2 seconds', 'Increase your speed by 5mph' or 'Shift lane in 5 seconds'. Such a system could only work if the rink master precisely knows the capabilities and desires of each participant, not to mention the unwieldy task of calculating the optimum decisions for hundreds of people and communicating it to them in time.

This idea directly ties with the centralized vs distributed decision debate in wireless communications. And from the skating rink analogy, we can see that if we want to accommodate a large number of devices with widely varying access capabilities and data rate or delay requirements, an open free-for-all approach can work well with even very limited guidance. The coincidence of interest in avoiding collisions and the statistical multiplexing of available data help create spontaneous order in the wireless domain.

What's essential for co-existence

There are two important things that minimizes accidents in the skating rink:

  1. A basic set of rules: 'No pushing or sudden stopping', 'No loose articles of clothing', 'Fixed direction of movement', etc.
  2. Visibility: More than anything, the reason why we don't see constant collisions is that everyone can see what others are doing and decide what's the best action to take in response.
In unlicensed band operation, we have an analogous set of rules that each device has to adhere to but sometimes the visibility of what others are doing is very limited. In particular what I think would really help promote co-existence is a basic sense of the wireless environment - more than what the device can see itself like what is the channel occupancy on other channels, what kinds of devices are operating nearby, how much of the channel are they occupying etc. Listen before talk and other MAC schemes on those lines, essentially make use of the local visibility and delay their transmissions based on what they see on the channel. However the end-result is far from optimum. Bianchi's analysis, for example, shows that there is a specific transmission probability at which the optimum value of the system throughput is achieved. And this transmission probability is a function of the number of participants in the interference range of a device, which is an unknown from the point of view of each device.

The final point

The key point that I want to convey here is that even though decentralized, distributed decisions seem to be the best solution for unlicensed operation, there could be great gains by providing each network/device a certain sense of the wireless neighborhood through external means (a control channel ?). Without an improved co-existence mechanism between secondary devices in the TV White Space, the whole ecosystem of this new unlicensed band could be in trouble.

2 comments:

  1. Interesting analogy.

    There is one distinction on my mind, which may or may not be relevant. On a public rink, being there is what a customer pays for, while in most communication systems, customers pay for a connection to the Internet, of which using the open band is a necessary part but not the only part. Customers pay an ISP for guaranteed QoS, which is not available on a public rink. (There might be a good startup opportunity for an ISP to provide cheap Internet connection without "hard" QoS)

    p.s. IEEE 802.19 (Wireless Coexistence Working Group) might be something similar to what you are pointing to in the last paragraph.

    ReplyDelete
  2. Thanks for the interest. Your point is definitely valid - from the economic point of view, the analogy would be with a toll road which you pay for and use, in order to reach somewhere. Though building a ISP business model for something which is free (WiFi) might be hard without the QoS guarantees.

    802.19 indeed addresses this topic and I have been following their work - hopefully we'll have a coexistence standard by next year in the white space band.

    ReplyDelete