On Thu, 18 Jul 2002 ggm(_at_)apnic(_dot_)net wrote:
A couple of general questions about 802.11 at IETF in Japan
1) is there some aspect of the telecommunications/regulations here
in Japan restricting the number of channels for RF use which affected
the problem? Did it help or hinder? -This is a 'cannot fix' problem
for IETF, but would be interesting to know, and to try and feed back
into national regulatory process for RF management. After all, the
density of endpoints and stations is probably close to a future world
dream.
actually japan has four non-overlapping channels 802.11 channel 1 6 11 14
whereas the US has only three because their 2.4ghz ism band goes from
2.4-2.5 and ours goes from 2.4-2.483. some commonwealth countries have
more stringent output regulations than the US or JP but that's not an
issue here.
2) is there some aspect of the density of wireless stations, station names
and SSID behaviour which contributed to the problem? I am hearing
gossip which suggests in some cases in RF space 'more is better' is
a false model, and that 3 stations in rooms where 4 are, might be
cleaner.
you need to do some engineering in order to make is such that the ap's
sitting on the same channels can't hear to much of each other. noise is
the biggest killer here because it results in more retransmissions which
results in deeper buffers on the ap's which results in more
retramismission which results in the ap's crashing from buffer-overflows
and overall results in reduced performance...
the accompaning issues is more than 60-100 clients per ap (and ~200=death)
really results in reduced performance as well, particulallry if most of
them are active so more ap's can result in better localized performance,
assuming you get a handle on the rf issue.
rogue ap and adhoc nodes result in their own set of confusion assuming you
client works right the biggest effect should be more noise, if your client
doesn't work properly (the apple one apparently) then other bad things may
happen.
A consideration here is that people choose to sit close to the
middle of the room, and some rooms the aerials are in the corners
so the area of maximum overlap of signal is ... the weighted centroid
of the middle of the room! Or the ven diagram overlap or whatever.
Maybe more 'randomy' placement of the stations, or all along one wall
would be better? I don't know. Could that kind of thing be tested
passively with an RF meter rather than with the massed laptops of IETF?
(and the hosting BCP updated...)
yeah it can be tested but the rf environement looks way wierder with 1200
addional transmitters and 50 tons of rf absorbing meat in the space then
it does without them so unless you have a very cool testbed environment
you don't really know what your changes will do unless you try them
And, the tricky third question
....or is this just yet another complexity which cannot be solved
by single-point changes...
cheers and thanks to the NOC who are clearly working very very hard to deliver
the best 802.11 they can, in trying circumstances (Typhoons do not help with
external aerials, rogue SSID ad-hoc, and who knows what else)
-George
--
George Michaelson | APNIC
Email: ggm(_at_)apnic(_dot_)net | PO Box 2131 Milton QLD 4064
Phone: +61 7 3858 3100 | Australia
Fax: +61 7 3858 3199 | http://www.apnic.net
--
--------------------------------------------------------------------------
Joel Jaeggli Academic User Services
joelja(_at_)darkwing(_dot_)uoregon(_dot_)edu
-- PGP Key Fingerprint: 1DE9 8FCA 51FB 4195 B42A 9C32 A30D 121E --
In Dr. Johnson's famous dictionary patriotism is defined as the last
resort of the scoundrel. With all due respect to an enlightened but
inferior lexicographer I beg to submit that it is the first.
-- Ambrose Bierce, "The Devil's Dictionary"