>Polling doesn't add overhead in the TC implementation.
that's an inaccurate statement
it actually does, when compared to stock 11b - that is why you turn off
polling for PtP links (e.g., b/n two satellites)
11b was designed for the LAN, with the intent of maximum data transfer
amongst a few stations
Polling is more advantageous than 11b in the situation when there are many
simultaneously transmitting stations off of one AP
e.g., the WISP scenario
>In TC, the more traffic on the network... the faster the network gets for
>everyone. Again, this because of polling combined with super-packets.
a fancy way of saying buffering -
super-packets exist in all polling implementation, as the Poller (AP) is
only capable of talking to one Client at a time
Clients waiting to be polled store traffic in a buffer, thus a super-packet
In terms of efficiency, super-packets are more efficient than 'standard'
packets mainly b/c there is less overhead involved in header info (one large
packet, vs. redundant header info of multiple smaller packets)
In terms of network speed increasing proportionally to the number of
clients - this is caused mainly by the buffering process of waiting stations
(w/ less clients, the stations don't wait as long, hence their buffers are
not as full, hence more waste b/c of smaller packets)
>Polling eliminates your packet collisions/hidden node problems, as the
>management of transmission/reception of signal is moved to the base
station.
true
>Most packets on the net are less than 75KB in size. So while polling other
>CPEs, if data is qued up for you at the A/P, the TC A/P doesn't wait to
send
>them 75KB at a time. Rather, it bunches them up in a big packet until it
is
>polled again or until it reaches about 2K in size. When it does send your
>packet, it also acks the previous transmission from another CPE (further
>saving time in the polling process).
basically, the station buffers while waiting to be polled
>Also, inactive CPE's are polled less often. Active CPEs are polled more
>often. So the guy who went to KFC for dinner is polled once a second,
while
>the guy downloading the latest nudy pictures from his favorite porn site is
>getting polled many times more often.
basically, a general overview on how dynamic polling works - what would be
nice would be for KarlNet to better explain the algorithm behind the
priority queuing - but then, that might be revealing a little too much <g>
-Charles
Doug Karl explained this so well at WISPCON. They should have made a video
of it and put it on the Karlnet web site. My explanation is much less
effective. Sorry.
Bill
_______________________________________________
Karlnet mailing list
Karlnet@WISPNotes.com
http://lists.wispnotes.com/mailman/listinfo/karlnet
|