<div dir='auto'><div>Hey Toke,<br><div class="gmail_extra"><br><div class="gmail_quote">Am 13.09.2018 21:12 schrieb Toke Høiland-Jørgensen <toke@toke.dk>:<br type="attribution"><blockquote class="quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><p dir="ltr">Ruben <ruben@vfn-nrw.de> writes:
<br>
<br>
> Hey Toke,
<br>
>
<br>
> Thanks for your fast response!
<br>
>
<br>
> Am 13.09.2018 12:27 schrieb Toke Høiland-Jørgensen <toke@toke.dk>:
<br>
>
<br>
>
<br>
> Ruben <ruben@vfn-nrw.de> writes:
<br>
>
<br>
> > Hey guys,
<br>
> >
<br>
> > I've already mentioned this in a response to dtaht on GitHub, but here
<br>
> > again for everyone:
<br>
> >
<br>
> > I was wondering if it's possible to extend the tin statistics by
<br>
> > packets for backlog.
<br>
>
<br>
> Why do you need packets when there's already bytes?
<br>
>
<br>
> Easy: dtaht requested a packets graph with ecn marks, which is also
<br>
> packets, so backlog as bytes do not fit, backlog as packets do.
<br>
>
<br>
> The idea was to do a multi-graph which is one graph with combined
<br>
> stats for all tins and sub graphs for all tins.
<br>
>
<br>
> On the main graph a backlog in packets is available, but I would need
<br>
> to leave out the backlog for the tins, which is somewhat confusing.
<br>
<br>
Why not just do both backlogs in bytes?
<br></p></blockquote></div></div></div><div dir="auto">There's no counter for ecn marked packets in bytes, so it's impossible to implement it that way, too.</div><div dir="auto"><br></div><div dir="auto">ECN-Marks is in packets, so everything else need to be in packets as well. </div><div dir="auto"><div class="gmail_extra"><div class="gmail_quote"><blockquote class="quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><p dir="ltr"><br>
> > Also I haven't found details on the man page, but haven't looked in
<br>
> > the code how exactly flows are defined for the new 2 level fair
<br>
> > queueings.
<br>
>
<br>
> A flow is a flow (hash on the regular 5-tuple). Unless you install
<br>
> custom tc filters, in which case a flow could be anything you want it to
<br>
> be...
<br>
>
<br>
> > I would like to show hosts as well as flows in a graph, in the various
<br>
> > states as summary as well as per tin. But hosts aren't shown in the
<br>
> > statistics. But they could be helpful to see unexpected rises if
<br>
> > someone is abusing the system.
<br>
>
<br>
> Hmm, guess we could do "number of hosts per tin" accounting as well as
<br>
> flows. But it would be bloating up the already bloated statistics. But
<br>
> on the other hand, they are already quite bloated, so another pair of
<br>
> stats maybe wouldn't make much of a difference?
<br>
>
<br>
>
<br>
> How important is a row more or less? It's quite verbose already (what
<br>
> I like) - as long as the number is available without much additionally
<br>
> work needed.
<br>
<br>
It's the additional complexity of recording the stats internally in cake
<br>
I'm worried about, not the extra lines of output :)
<br></p></blockquote></div></div></div><div dir="auto"><br></div><div dir="auto">Alright.</div><div dir="auto"><br></div><div dir="auto"><br></div><div dir="auto">Best regards</div><div dir="auto"><br></div><div dir="auto"><br></div><div dir="auto">Ruben</div><div dir="auto"><div class="gmail_extra"><div class="gmail_quote"><blockquote class="quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><p dir="ltr">
</p>
</blockquote></div><br></div></div></div>