All of the issues listed here are great examples of the complex nature
of designing an 802.11 environment with such stringent requirements.
With only 3 channels, even if you plan very carefully and precisely
control the output power of your APs, you're going to get channel
overlap. This will further reduce your capacity due to the inherent
collisions/retransmissions. Especially when you factor in the client
devices. A client device transmitting on a channel will force any other
device operating on the same channel that can hear it (APs included if
course) to wait on it to complete its transmission before it can
commence. So, you have to realize that, even though 2 APs may not be
able to hear each other, a client card between them that can hear both
of them will tie up available bandwidth on BOTH APs while it is
transmitting. Further complicating matters is a situation where two
clients connected to two different APs on the same channel can hear each
other but not both APs. In such a circumstance, client 1 and the AP 2
(the AP client 2 is connected) may transmit simultaneously. When this
happens the signals will interfere with each other upon reaching client
2, causing client 2 to be unable to decode the packet, forcing AP 2 to
retransmit the packet.
Complicated indeed! Guaranteeing signal strengh and bandwidth alotments
is extremely difficult. And, this totally ignores the problems inherent
with outside interference or the fact that the environment (bookshelves,
etc) change on a regular basis, possibly forcing you to revisit your
ever-so-finely-tuned RF plan. Interestingly enough, all these issues
are also extremely relevant if you're interested in looking to deploy
any sort of VoIP/WiFi (VoFi).
I'd suggest that, if you're truly interested in providing
coverage/bandwidth that takes a lot of these issues into account, you
might want to take a look at the Meru Virtual AP architecture. The
controllers in these systems keep track of every 802.11 device each AP
can here and employ a pretty darn impressive scheduling algorithm for
getting the most out of the available channel capacity. Not only that,
but they actually control when clients are allowed to transmit, further
removing unknowns from the RF use equations and improving channel usage
and capacity. I believe they do this using the PCF, or Point
Coordination Function, in the 802.11 spec... I've not seen any other
wireless switch system that makes use of it near to the level that the
Meru system does. It's pretty cool. We're in the process of deploying
Meru as our second generation wireless overlay here at UTD, mainly to
decrease the need for complex channel planning, individual AP
configuration, and to support a future VoFi implementation.
--Mike
Phil Raymond wrote:
If someone forced me to assign a rule of thumb at this high level, I
would assign a conservative data rate of 1 Mbps to each student as a
requirement. For an 802.11g ONLY network running at the highest data
rate (aka strongest signal) using enterprise class AP's (data thruput
does vary between AP vendors, be careful here), you should expect to get
15-20 Mbps of upper layer thruput per AP. That would yield 15-20
students per AP. For 802.11a, this will probably hold. For 802.11g, due
to the limit of 3 channels, you will get an overall reduction in
capacity due to shared bandwidth between AP's in a densely deployed AP
environment.
Also, this assumes that you design the network for the highest signal
strength - a very important point. In most instances this won't be
possible due to the environment. Thus I would reduce the available
bandwidth by 33% and say that 10Mbps is available.
Hence I would go with the low end of 10Mbps available per AP.
To take this to a lower level of analysis, I would want to know what
applications the students would be running. Perhaps you use the analogy
of a low end DSL connection that provides 768Kbps downlink and 128kbps
uplink. Then you stick with the 1 Mbps/student and assume it supports
most if not all applications they will use. You might also consider a
swag at peak operating times (evenings) and assume ~50% of the available
students are online (simple queuing theory assumption). Then you could
say that a single AP would cover minimally 20 students. There is my rule
of thumb at this high level. I would consider it conservative if you
design the network properly.
In a typical dorm with a lot of walls (and bookcases...), you will
probably find that your coverage requirements and capacity requirements
will be in alignment (and thus balanced). What I mean by that is that
you will find that in order to provide a good signal in a dorm
environment you will need to place a denser AP deployment (due to the
thick walls, etc.). This means that as a consequence your capacity will
also be increased due to the denser deployment.
Other factors not considered here are the use of client cards.
Performance between different manufacturers (you get what you pay for)
will vary. Some cards will be noisy and interfere, others will have
higher SNR requirements, etc.
Hope this helps and not confuses - as I said, it is not a trivial
subject.
-----Original Message-----
From: Larry Press [mailto:[EMAIL PROTECTED]
Sent: Wednesday, November 09, 2005 9:51 AM
To: WIRELESS-LAN@LISTSERV.EDUCAUSE.EDU
Subject: Re: [WIRELESS-LAN] Wireless-only Dorms?
Phil Raymond wrote:
The initial design needs to consider coverage AND capacity.
Phil (and others),
Have you got a rule of thumb for the number of students per G access
point
in a college dorm?
Larry Press
**********
Participation and subscription information for this EDUCAUSE Constituent
Group discussion list can be found at http://www.educause.edu/groups/.
**********
Participation and subscription information for this EDUCAUSE Constituent Group
discussion list can be found at http://www.educause.edu/groups/.
**********
Participation and subscription information for this EDUCAUSE Constituent Group
discussion list can be found at http://www.educause.edu/groups/.
