From: Erick B. (erickbe@xxxxxxxxx)
Date: Tue Sep 05 2000 - 02:39:12 GMT-3
I've seen out-of-sequence problems with
equal-cost-routing between different vendors boxes. It
boils down to how the box routes over equal-cost
routes and how it's configured (route-cache, no
route-cache, cef, etc). I think out-of-sequence
packets is likely w/equal-cost routing even between
same vendor boxes. The switching methods are local to
each router/networking device so one side may forward
x traffic over one link, and the other box may forward
it over the other link, and which is sent to the end
device first?? It could be a end-device/application
issue of handling packets as well. This isn't a area
I'm a expert in so...
I believe combining ports into 1 channel/group/trunk
would be better if going between same-vendors box,
therefore its a big pipe with 1 route and you have
less chance of out-of-sequence packets.
--- damien <damien@clara.co.uk> wrote:
>
> Has anybody actually seen problems with out of
> sequence packets in real life scenarios. The
> scenario I am interested in is the typical example
> where there are switch paths to the same destination
> with one link for example 2M and the other 1M? Does
> anybody have any proof of concept testing done in a
> Lab were they can confirm at what different speed
> ratios you are sure to experience problems based on
> certain types of switching etc? I appreciate that
> there are a lot of variables in this equation, e.g.
> type of switching, size of traffic streams etc, etc.
> Any information/ experiences would be appreciated.
>
> Can Variance overcome this issue ? Without reviewing
> my notes, I can not remember whether Variance
> switches on stream or packet. If it is stream. In
> theory, if a Router is performing stream switching,
> is it fair to say that there should be no out of
> sequence issues due to the fact that the entire
> stream is following a single path with fixed speeds.
> If per packet switching is in use, they there maybe
> issues with out of sequence packets due to the
> packets belonging to the same stream being switched
> via different speed links.
>
> I could go on an on........ :~)
>
This archive was generated by hypermail 2.1.4 : Thu Jun 13 2002 - 08:24:52 GMT-3