On Sun, 15 Dec 2002, Kevin Knuth wrote:
> The TurboCell algorithm is pretty complex, so I am not sure of the "delay or
> timeout" before it moves on.
>
> However, you must realize the forcing TurboCell to "not wait" has an effect
> too! I just wonder if the benefit outweighs the other possible effects on
> performance. A question that WISPs have to ask themselves is how many links
> do they have that have less then great signal? Most WISPs run with very
> marginal links. If we tell TurboCell to ignore those, they may as well close
> up shop!
Hello,
As I said, this problem only occurs for a few minutes, where a single
satellite is hogging all resources. No other satellites get polled. I
don't believe that this "fix" would necessarily have to take into account
a constant situation but rather a particular case where the satellite is
obviously not sending something back to the base and it gets stuck. I
don't see how it can be considered normal that a satellite can cause an
entire multipoint setup to become unuseable for a number of minutes.
As for marginal, during one of these "problems", I had 20db+ SNR on either
the base or the satellite, therefore I doubt it is a noise issue. And the
satellites in question have less than 2% retransmits at any given time
other than these particular incidents where we have several thousand % of
retransmit.
Chris
|