Wireless Central Coordinated Protocol

[Darryl Smith]

People

This is what I love - a good debate as to some of the technology behind
the technology... In researching my initial reply I certainly did learn
some things about 802.11b, and how it operates. 

The reason why I believe that the correct solution is to use power
control is the fact that Spread Spectrum (SS) requires power control to
operate correctly. The most evident example of this is the 'Near-Far'
problem that you can find described in literature, which is very similar
to the capture effect in FM. It is differing power levels that are the
issue with the tendancy of nearer clients to get more access to the AP
than those further away.

Until I examined the documentation on the protocol I did not know about
this problem with the 802.11b MACA protocol, but it makes sense. MACA
does not solve all problems, it just reduces them... Being an electrical
engineer, I like to start with issues at layer one of the 7 layer stack
- the physical layer. Once the bottom layer is fixed, the other layers
can be worked on. Protocols can only work if the foundations are there.

Abstraction to IEEE Model

>From what I have read of the proposal you have gone from basically a
CSMA/CD model, as per 802.3 for the MAC, to 802.4 Token Bus. I must
admit that this does have many advantages - particularly where traffic
must get through, and in the situation where the traffic flows are
predicatable - or not bursty. For those that are not familiar with the
lower levels of wired ethernet, Token Bus is CoAx Ethernet, but with a
token passed between stations informing which can talk when. 

I think it is probably wise that you have not implemented 802.4 - as
Tanenbaum notes in "Computer Networks", 'The 802.4 MAC is very complex,
with each station maintaining 10 different timers and more than two
dozen internal state variables'. Actually Token Bus is not exactly what
is being proposed, since you are proposing that there be a permament
master I believe... At least that is how I would do it.

When WiCCP works and when it does not

It is interesting to note under what conditions WiCCP will work, and
when it will not work - at least optimally. WiCCP will out perform
systems that do not run it when the utilisation of the bandwidth
increases above some high percentage. If we were running standard
ethernet the figure would be about 80%. I have no idea what it would be
in this case. Above this figure, whatever that is, the channel
assignment ability of WiCCP will allow the utilization to increase
almost to 100% - or at least as close as is humanly possible.

Looking on the other end of the scale, standard 802.11b will work best
when the utilisation is low, and the levels are set correctly so that at
the access point all power level are the same. Under low utilization it
is likely that the power levels do not affect things too much.

The main question regards bursty traffic. WiCCP allows a guarentee of
bandwidth for a particular user, and this solution appears to be the
correct solution for this case to solve this problem. The ability to
offer a guarenttee and then offer more on top of that where available is
worthwhile. The problem is the overhead of polling.

As the number of users increases, WiCCP will tend to have issues with
assigning timeslots to each, ensuring latency. Standard 802.11b will
have a definite advantage whem there are a lot of stations, and very few
want to transmit most of the time - Such as 50 laptops who only check
their mail once every 15 minutes [but not to read it], as oposed to 50
users attempting to surf the web.


Gus:
Have you investigated the effect of Short Vs. Long preambles on normal
802.11b? I think that you might find that with short preambles the
period that RTS/CTS packets can be corrupted is significantly smaller,
and therefore MACA is more likely to work. The only reason why RTS/CTS
works is that those packets are 'significantly' smaller than the normal
payload packets, and therefore more likely to get through. 

I think that is enough from me, else I will find the message size limit
on this mailing list

---------
Darryl Smith, VK2TDS   POBox 169 Ingleburn NSW 2565 Australia
Mobile Number 0412 929 634 [+61 4 12 929 634 International] 
Darryl at radio-active.net.au | www.radio-active.net.au