I think my experience with the wireless network was on the lower end
of the scale, perhaps in part due to the fact that I am using an older
laptop and Cisco 340 card with 802.11b only. I also run FreeBSD, so
that put me outside the norms. The wireless network worked great on
Saturday and most of Sunday, but ran into trouble as the population
grew.
As a consequence, I spent some time working with the NOC folks after
they invited assistance to investigate problem situations. They
explained to me that the equipment used in this network has dumb
access points hooked to centralized controllers. Those controllers
implement or proxy the DHCP and ARP functions. Furthermore, DHCP, ARP
and data packet forwarding are separate functional paths through the
controller which can independently work better or worse. That can
result in surprising behavior relative to a more traditional
functional decomposition into separate boxes. For example, a number
of times I got an address from DHCP, but could not ARP the first-hop
router. Some of the improvements over the days came from turning off
features such as one that tries to protect against address use by
sources that had not obtained an address via DHCP. In the later days,
I still had problems becomming disassociated, but always when
associated I got DHCP, ARP and packet flow.
I noticed that performance was always fine once the session was over
and most people shut down their connections. One hypothesis was that
limits were reached in various state tables in the controller. Only
the vendor could know about those details for sure. But I draw the
conclusion that this particular system may not be ready yet to scale
to the density of radio participants at the IETF meeting.
-- Steve
_______________________________________________
Ietf@xxxxxxxx
https://www1.ietf.org/mailman/listinfo/ietf