Thanks for your posting.
The packet traces of both Cisco and Meru are downloadable here:
http://i.cmpnet.com/NetworkComputing/design06/downloads/mobile/captures.zip
A couple of months ago I exported the traces into a text file, extracted the
duration value, and built histograms of them and compared them between Cisco
and Meru. While there were differences, nothing really jumped out at me. I
also spent some time reading the 802.11-1999 specification and dug in to the
duration value descriptions. There's not a static calculation to derive the
duration value -- there are dynamic things going on in the air that will
feed into the AP assigning it a certain value. I don't think it's possible
to take the trace file and say that the values should have been definitively
this, that, or another. The Meru values didn't seem grossly off-base.
Russ' example test case about co-locating three microcell APs on different,
non-overlapping channels and comparing that to three virtual cell APs on the
same channel will likely prove that the microcells will win in that
scenario. What Meru and Extricom are saying (and even Xirrus) is that
performance drop due to co-channel interference of neighboring APs in a
microcell configuration is bad enough that you're almost reducing it to a
single-channel configuration, but with all the additional backoffs and
retries. I don't think we should under-estimate the negative effect of
co-channel interference in micro-cell architectures where there are high
throughput and wireless QoS demands. Again, it won't be an issue with in
low-volume wireless networks, but I think it will be exacerbated as your
students drive the traffic volumes up.
Frank
_
From: Russ Leathe [mailto:[EMAIL PROTECTED]
Sent: Thursday, April 12, 2007 9:54 AM
To: [EMAIL PROTECTED]
Subject: [WIRELESS-LAN] Re : microcell vs virtual cell
Hi Everyone,
Just wanted to weigh in on this conversation.
We went through this same process at Gordon College about 4 years ago. We
looked at Meru, Trapeze, AirSpace and Aruba. During this process, there
were no independent studies or analysis of anyone's product.
Gordon College pilot's any new technology. Our pilots run between 30-60
days.
We also rely on analysis by Network Computing, Gartner and Tolly.
Typically, they do a very good, non-bias, review of products and technology.
Next we look at the company and the amount of $$ they spend on RD, patent
ownership and whether or not them have engineers or a presence on standards
committee's. It's the standards that really catch my eye. Hopefully, if
they conform, than I have positioned myself and the institution for the
future. If not, then the trouble begins.
Needless to say, we chose Alcatel/Aruba. They conform to standards. So
far, we have rolled out WifiVOIP, Multicast (video/audio), wifi VPN,
seamless roaming. It all works.
I read the article regarding Network Computing analysis of CISCO vs. Meru.
Honestly, they did a very thorough job. What I got out of the article was
that indeed Meru was/is tweaking the 802.11 duration value. To me, this is
a red flag. This means they are not conforming to standards. Has Meru
responded to Network World or their customers? Seems to me if this wasn't
true, there would be lawsuits or at least a rebuttal. Here's one of the
engineers e-mail address, [EMAIL PROTECTED] I would contact him and ask the
question.
Anyway, let's get back to testing or piloting technology.
By virtual cell vs traditional micro-cell WiFi I assume you are talking
about having multiple access points advertise the same BSSID on the same
channel. Virtual cell appears to clients as a single access point. In this
definition virtual cell would result in more clients contending for the
given channel within a larger coverage area.
Contention in WiFi is generally managed by CSMA/CA (collision avoidance). CA
is used instead of CD (collision detection) because clients are sometimes
not within range of other clients. You can create virtual-contention to go
with the virtual-cells by tweaking access point timers to be shorter than
client timers, but the single channel or access point bandwidth is fixed.
You raise a very good question about the data/analysis since the number of
vendors promoting such concepts is very limited.
Place 3 micro-cell access points on 3 different channels in a coverage area.
Associate 3 clients 1 per access point. Measure. Place 3 virtual-cell access
points on same channel in same coverage area.
Associate 3 clients. Measure.
Be sure to take bi-directional measurements with simultaneous TX/RX to
experience the half-duplex nature of WiFi radio. I am confident you will
push more packets over 3 channels than 1. Be sure to power-off the system
you are not testing. Shorting timers or counters by one system can adversely
impact other client and access point devices even with only background
traffic.
There is 3rd party vendor/bake-off documenting
