Lee,

I'm at 5.2.178.16 - That controller code adds support for disabling world-mode 
IE on a AP or controller-wide basis.

Jeff

>>> Lee H Badman <[email protected]> 5/13/2009 8:44 AM >>>
Great info, Jeff. All switch ports are up to snuff- again, am just firming up 
for controlled testing. But the World Mode IE is a bit of a question- am 
checking with both vendors as is not obvious in the interface.  Can you remind 
me what code/APs you're running?

-Lee

Lee H. Badman
Wireless/Network Engineer
Information Technology and Services
Syracuse University
315 443-3003

-----Original Message-----
From: Jeffrey Sessler [mailto:[email protected]] 
Sent: Wednesday, May 13, 2009 11:27 AM
To: [email protected]; Lee H Badman
Subject: Re: [WIRELESS-LAN] Sort of a fuzzy question on 11n

Lee,

Since you mention the Mac, most are broadcom, and I'm aware of a serious bug 
with broadcom chipsets that do cause wild swings in the data rates. I've only 
observed this on 5GHz.

The issue is with any AP that includes the World Mode IE (information element). 
When I broadcom-based client sees this, it will adjust it's power output to 
zero, back up, down, repeat and rinse. If the vendor supports a way to disable 
world mode IE, I'd do so and retest. Cisco built special radio code for us that 
disabled it while we wait for broadcom to fix the driver issue. Cisco was able 
to duplicate this on several broadcom-based computers including the Macs and 
HP. So, check those clients with the wild swings to see if they are 
broadcom-based.

I have several test clients here (broadcom and atheros) that are getting 
consistent (repeatable) symmetric data rates in the 110-120Mb/s from several 
different client types. The atheros-based is always better than the broadcom. 
There are ways "in the lab" to get these rates at or above 150Mb/s, but you now 
have a single-client race-horse so not really geared for typical "real-world" 
use. 

Intel-based clients are hit-miss depending on the driver versions. Intel seems 
to have some issues over the past year, but the latest driver seems to be much 
better. That said, I do know that I've seen more asymmetrical rates (download 
slower than up) on the Intel.

What's the lab setup look like? Is everything connected via gig/10gig? Is the 
tool you're testing with local to the same switch that controllers/AP's are on?

Oh, one last test... If you force your in-house test system to use a 100Mb/s FD 
connection, do the client rates become symmetrical and consistent? Have you 
tried testing using the I2 NDT test tool?

Jeff

 

>>> Lee H Badman <[email protected]> 05/13/09 5:55 AM >>>
This is not a vendor invite to "educate" me. Please, no vendor calls.

As we play with a couple of the major players' 11n stuff (and a couple of 
consumer grade units for giggles)- am seeing a definite trend. Wondering if 
others have seen the same. I know of the many variables of wireless in general, 
and specifically with 11n, but despite all of that I'm still a bit surprised 
over these.

The quick setup:

-          wide channels only
-          5 GHz only, 2.4 radios disabled
-          Channels are not interfering between hardware sets
-          Where it can be toggled in hardware, SGI is in use
-          All clients tested close to each AP, and with a wall or two between
-          2 X 3 and 3 X 3 in use
-          Variety of clients- XP/Mac/Vista on WPA2/EAS/PEAP-MSCHAPv2- no-NAC 
connections
-          In-house speed test utilities and FTP transfers used in early testing

What seems strange, in general:

-          The wild inconsistency of data rates. On the new Mac versions you 
can see data rate, channel, associated AP by doing "alt- click on radar icon". 
If you do this repeatedly from several different Macs, they all show data rates 
that fluctuate between 300 Mbps and very low rates- like 6 Mbps. Doesn't matter 
of you have straight on LOS or walls and multipath, and regardless of what 
hardware set you're using, this fluctuation can be shown at will
-          Actual throughput in the 11a/11g world is typically slightly under 
half of the stated date rate in conversation and function. I'm seeing 11n is 
more like 1/3 (or less)- if client data rate is stated as 300 Mbps, throughput 
is at best around 110 (or 100, or 90) on interference free channels- regardless 
of hardware set
-          From the same spot, download/upload speeds can very wildly- test 
after test might be different (stated data rate of 300, typical throughputs of 
60 Mbps up/90 down, 110/ 80, 70/75, 113/90, 60/65, 100,58, etc) from same 
client, and all clients tested act the same way- wildly variable

I don't consider any of this "controlled" testing yet, and am just now starting 
to get ready to do for-real testing as we slowly advance toward 11n. But wanted 
to float these observations and see if others have found similar variability 
and have drawn conclusions about it.

-Lee

Lee H. Badman
Wireless/Network Engineer
Information Technology and Services
Syracuse University
315 443-3003


**********
Participation and subscription information for this EDUCAUSE Constituent Group 
discussion list can be found at http://www.educause.edu/groups/.

</[email protected]>

**********
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/.

Reply via email to