InternetMCI(purchased by Cable & Wireless) and AT&T spent a lot of
effort developing their tools. I haven't seen any commercially available
systems which come close. C&W and AT&T seem to be unique among huge backbone
providers in publishing their network performance.
However, an alternative for measuring round-trip times with ICMP is
setting up a full mesh of NTP assocations. As part of the calculations
of the "time" get an ongoing measurement of network delay between the
routers. You can use your favorite tool to read the NTP associations
table, and generate a nice web page. I believe this is how UUNET use
to do their version.
> > What tools exist that will periodically log into multiple routers and run
> > ping tests
> > to various destinations and then record the results in graphical form?
>
> maybe the RTR cisco router feature for sonde-like probing ?
InternetMCI(purchased by Cable & Wireless) and AT&T spent a lot of
effort developing their tools. I haven't seen any commercially available
systems which come close. C&W and AT&T seem to be unique among huge backbone
providers in publishing their network performance.
However, an alternative for measuring round-trip times with ICMP is
setting up a full mesh of NTP assocations. As part of the calculations
of the "time" get an ongoing measurement of network delay between the
routers. You can use your favorite tool to read the NTP associations
table, and generate a nice web page. I believe this is how UUNET use
to do their version.
There used to be a site somewhere (I think Andover or Keynote sucked up
the domain, but it was a while ago) that had a very nice little script
that pinged IP addresses and presented the visitor with average latency
to those addresses. Looked something like this:
UUNet (x.x.x.x) -> 38 ms (green)
C&W (x.x.x.x) -> 68 ms (yellow)
AOL (x.x.x.x) -> 400 ms (red)
It was cheesy, and not particularly scientific, but I've been trying to
find something like that to implement for the marketing folk. It could
probably be adapted into something more useful to us though. Suffice it
to say, I haven't found anything, and I'm about to dive into a week-long
coding spree writing my own. If anyone can connect me with something
that'll work, I'd be happy to send a case of their choice of beer.
There used to be a site somewhere (I think Andover or Keynote sucked up
the domain, but it was a while ago) that had a very nice little script
that pinged IP addresses and presented the visitor with average latency
to those addresses. Looked something like this:
UUNet (x.x.x.x) -> 38 ms (green)
C&W (x.x.x.x) -> 68 ms (yellow)
AOL (x.x.x.x) -> 400 ms (red)
It was cheesy, and not particularly scientific, but I've been trying to
find something like that to implement for the marketing folk. It could
probably be adapted into something more useful to us though. Suffice it
fping, from Stanford originally, now at www.fping.com
might be useful, it pings multiple hosts at the same time
(fast, efficient) It has easy to parse output and easily gives results
like:
it's pretty cool stuff. requires *nix box, perl5, and some sort of
webserver software to produce simple reporting. there's also (optionally)
utilities that draw some pretty graphs that require gnuplot/ppmtogif.
imho, this is considerably better than logging into your router to do this.
routers are much better at forwarding packets than sending/receiving
them. (except older non-distributed routers, which aren't particularly
great at either for high traffic volumes) other bonus: no automated sending
of passwords from a box that might not get much admin attention.
one could probably modify these tools to use fping, but i just played
around with them for edutainment purposes. there's no mention of copyright
that i can find, but one should ask before using for commercial purposes.
I once worked for a company that wrote a unix script that worked like this.
Basically imagine a quare chart will all the pops listed across the top and
down the left side. Every few minutes, each pop tries a small ping burst to
ping all of the others, and the values are filled into the chart. Results
are color coded as green, yellow, and red.
I once worked for a company that wrote a unix script that worked like this.
Basically imagine a quare chart will all the pops listed across the top and
down the left side. Every few minutes, each pop tries a small ping burst to
ping all of the others, and the values are filled into the chart. Results
are color coded as green, yellow, and red.
>
> when someone asked me to do something like this, i waded through caida's
> site and came accross this:
>
> http://www-iepm.slac.stanford.edu/pinger/
>
> it's pretty cool stuff. requires *nix box, perl5, and some sort of
> webserver software to produce simple reporting. there's also (optionally)
> utilities that draw some pretty graphs that require gnuplot/ppmtogif.
>
> imho, this is considerably better than logging into your router to do
this.
> routers are much better at forwarding packets than sending/receiving
> them. (except older non-distributed routers, which aren't particularly
> great at either for high traffic volumes) other bonus: no automated
sending
> of passwords from a box that might not get much admin attention.
>
> one could probably modify these tools to use fping, but i just played
> around with them for edutainment purposes. there's no mention of copyright
> that i can find, but one should ask before using for commercial purposes.
>
> >
> > > It was cheesy, and not particularly scientific, but I've been trying
to
> > > find something like that to implement for the marketing folk. It could
> > > probably be adapted into something more useful to us though. Suffice
it
> >
> > fping, from Stanford originally, now at www.fping.com
> > might be useful, it pings multiple hosts at the same time
> > (fast, efficient) It has easy to parse output and easily gives results
> > like:
> >
> > fping -e <targets
> > www.chatt.net is alive (0.32 ms)
> > www.att.net is alive (27.5 ms)
> > www.uu.net is unreachable
>
> --
> Sam Thomas
> Geek Mercenary
>
The NLANR Multicast beacon does this now for multicast :
It would be pretty trivial to modify this for a set of unicast beacons; the code is available.
There would have to be beacon management or discovery (now done implicitly by multicast).
You might want to lower the beacon send rate as well.
Both C&W and Digex/Intermedia had websites that showed this information.
However, this was only shows within that provider's network, not to other
networks.
