Error Message
Fairlight
fairlite at fairlite.com
Fri Nov 30 08:58:21 PST 2007
On Fri, Nov 30, 2007 at 11:26:29AM -0500, Jay Ashworth may or may not have
proven themselves an utter git by pronouncing:
> > Which is almost useless as they include no sequence number. If 99% of your
> > times are all 0.1ms, you can't tell squat once you've filled the screen and
> > scrollback buffer to min_scrollbar_size.
>
> For our purposes, which are mostly "what percentage of packets are
> being dropped on this link", what do you need the seq numbers for,
> Mark?
Because that's not how I'd be utilising it, were it me.
When I have a question about correlative events in networking, if I'm
testing connectivity with ping, I'll let it run continuously from *nix.
If there are gaps in realtime in the sequence numbers -while- a "network
event" (ie., problem) is occurring, I can see immediately that it's not
just a general loss, it's a direct correlation at that particular point in
time; there is a direct and immediate correlative match.
If you just let Windows' ping run endlessly, have an event, and break ping
later for your report, there's no guarantee of direct correlation because
you've not observed one at the same time (due to inability through lack of
SEQ), unless you get lucky and get a variance in latency that happens to
show up during the outage and lets you know that the thing is still alive.
Breaking it at the end and seeing "some" loss doesn't tell you -when- the
loss occurred. Hence, no correlation, and far less substantial proof that
there's a connection (or not) to the event being diagnosed. All it tells
you is that there was an issue at some point, but not that there was an
issue -during that event-.
mark->
More information about the Filepro-list
mailing list