Hi,
The main drawback I'm seeing with cacti is that it is triggered by a
cron job. The resulting load of the collector looks like a square wave:
+-----+ +-----+ +--
| | | | |
| | | | |
---+ +---+ +---+
| |
|<------->|
5min
The other point is that I did not find how to control the degree of
parallelism of the collector. So is some routers are answering too
slowly, the whole process is slowed down and the system may not be able
to complete the cycle in the 5 minutes. This may be OK for a couple of
interfaces, but with many thousands of them, it becomes more
challenging.
So this is why we have developed our own system. It runs as a daemon
with its own internal scheduler and is able to poll many routers in
parallel. Our scheduler is implemented using a priority queue and the
resulting load of the polling server is almost constant.
The other problem with Cacti is the need of an additional mysql
database. We already have a database that contains all the information
needed for the polling, so we did not need to duplicate this information
in yet another database.
But don't get me wrong, I like Cacti and Cacti is great for monitoring a
dozen of servers or a couple of hundred routers. On our platform, it
just did not scale for polling many thousands of interface and its
integration within our current database infrastructure was not
straightforward.
Stanislav gave us an interesting link to rrfw. I have to admit that I
don't know this tools, but it doesn't look too bad (at least on the
feature list published on the web). However, I did not find anything
about parallelism on the web.
Cheers,
-- Jacques.
On mer, 2004-07-21 at 16:29, Frederic Gargula wrote:
> Jacques,
>
> Yes, I know it's RRD generated (yes it looks like all the same)...
>
> You could link to rrdtool.org to make Tobi happy :-)
> (see http://people.ee.ethz.ch/~oetiker/webtools/rrdtool/license.html)
> (Ok, with a rrdtool logo, it wouldn't match anymore your website
> graphical line ;-) )
> On our side, we leave the little "RRDTOOL / TOBI OETIKER" text in the
> upper right corner of generated images"
>
> Don't take me wrong about your graphs.. It wasn't a feedback about the
> colors, but much more about the graphs presentation.
> Was I awaiting for a beautiful stats website from IP-Plus ? Surely ;)
>
> "Ugly" was humoristic, probably to make you react.. What you did :-)
>
> Cacti matches our needs, and I don't know any other software with all
> those features and possibilities.. (I think about CDEF, stack graphs or
> even gradient graphs... Ok you can do that with your own tool, but wheel
> exists already)
>
> And the technical point come, which may interest swinogers.. Why do you
> say that cacti is "way too inefficient" ? What do Swinogers think about
> it ? Who's using it ?
>
> Anyway, thanks for your comments.. And as you said, let's keep the Swiss
> Internet community on a *constructive* track!
>
> Fred Gargula
> SIG / IP-Man
>
>
> Jacques Supcik wrote:
> > Dear Frederic,
> >
> > We have just migrated our statistics to a new system and we are very
> > happy to receive constructive feedback from our customers and from the
> > Internet community.
> >
> > The peak has been removed and I've corrected the percentage, thank you
> > for feedback.
> >
> > About Cacti: yes, we have indeed tried Cacti and we are even using it
> > internally. For our needs Cacti is way too inefficient, so we have
> > developed our own system that uses RRDs directly. The only differences
> > with the graphs that we have on our web page and those produces by a
> > "native" Cacti are the color and the texts at the bottom of the graph.
> > (Please note that both graphs are produced by RRD!). You may prefer the
> > green versus the blue, but I don't think that this allows you to qualify
> > our graphs as "ugly". Anyhow, the colors that we are using are
> > consistent with the rest of our web site.
> >
> >
> > Thank you again for your contribution and keep on giving *constructive*
> > feedbacks on this mailing list.
> >
> > --
> > Dr. Jacques Supcik
> >
> > Swisscom Enterprise Solutions AG
> > Network Applications Development
> > IP-Plus Internet Services
> > Genfergasse 14
> > Postfach
> > CH-3050 Bern
> >
> > E-Mail: [EMAIL PROTECTED]
> > Internet: http: //www.swisscom.com/enterprise-solutions
> >
> >
> >
> > On mer, 2004-07-21 at 14:35, Frederic Gargula wrote:
> >
> >>Hi all,
> >>
> >>Roger Schmid wrote:
> >>
> >>>Hi all,
> >>>http://www.ip-plus.net/network/traffic_load/index.en.mpl?SYS_state=init&SYS_nav_lnk=zhtix&SYS_pager_go=
> >>>interresting interface, would know which technology could made such
> >>>speed happen
> >>>
> >>>amused ..
> >>
> >>What about the percentages ? :-)
> >>And do they use MRTG to produce those ugly graphs ?
> >> [So, none of them tried Cacti ? (www.raxnet.net)]
> >>
> >>_______________________________________________
> >>swinog mailing list
> >>[EMAIL PROTECTED]
> >>http://lists.init7.net/cgi-bin/mailman/listinfo/swinog
> >
> >
> > _______________________________________________
> > swinog mailing list
> > [EMAIL PROTECTED]
> > http://lists.init7.net/cgi-bin/mailman/listinfo/swinog
>
> _______________________________________________
> swinog mailing list
> [EMAIL PROTECTED]
> http://lists.init7.net/cgi-bin/mailman/listinfo/swinog
_______________________________________________
swinog mailing list
[EMAIL PROTECTED]
http://lists.init7.net/cgi-bin/mailman/listinfo/swinog
