Re: CAPWAP BOF follow up: nmrg - CAPWAP

[Randy Bush <randy@psg.com>]

>>> Secure boot would be a useful thing to have
>> but how far does that go toward allowing me to deploy a mixed
>> vendor wireless network?
> Secure boot is an orthogonal issue.  It's what happens *after*
> the boot imagine is loaded that is important for the purposes of
> interoperability.

== secure red herring

>> i believe that a bit of thought will make you suspect that a
>> multi-device environment, where different devices have different
>> load-balancing algorithms, or even the same algorithms with
>> different constants, will lead to serious thrashing.
> We see thrashing already even where the devices have the same
> algorithm.  For example one vendor who shall remain nameless
> exports the number of users on the AP.  Unfortunately, this
> causes the clients to associate with an AP that has recently
> rebooted -- which is often the one that has got something wrong
> with it.  This kind of instability can occur with any "load"
> based load balancing algorithm if an AP has a tendency to crash
> or reboot under high load.

has anyone suggested an algorithm which performs reasonably in such
circumstances?

> To my knowledge, I'm not aware of any implemented products that
> try to determine whether an AP is actually functioning before
> moving client load to it.

what better test of functionality than to ship the product and see
if the customer complains.

> Given that effective load balancing requires interactions between
> clients and APs, and given the quality of vendor "value add" so
> far, it would be difficult for the IETF to create algorithms any
> worse than what is already shipping in today's products.

i have faith that, with a bit of effort, we can.

randy