[TriLUG] Request for help: residential internet service provider options.

[Steve Holton via TriLUG]

Excellent suggestions, Thanks!!

I'll need to review my options when I get home.

One thing I have noticed is that the PPPOE link is repeatedly dying. The
2nd level at Windstream said I had lost link 30 times over 3 days, which is
roughly consistent with what syslog reports.  The scenario seems to be that
if 5 sequential pings over the PPPOE link fail, the connection is
automatically reset.   I'll look into this again when I get home.

Thanks again.


On Wed, May 25, 2016 at 11:35 AM, Matt Flyer <matt at noway2.thruhere.net>
wrote:

> Good Question,
>
> Most of the quality tests seem to be oriented on one of a handful of
> testing sights for upload, download, and ping, which may not give you the
> best results; like you seem to be experiencing with netalyzer.
>
> The first one that comes to mind is MTR, which should show you what node
> hops are slow and perhaps let you pinpoint the exact problem spot.
>
> Other things that come to mind include using WGET to try to download a
> large file and see what the transfer rate is or if it speeds up and slows
> down or even stalls.  You could run this at various times to compare.
>
> Have you looked at the diagnostic output for your network interfaces?  For
> example, are they showing a large packet drop or large number of errors?
>
> Are there any services you could connect to that would remain active and
> inform you if you have a dropped connection?  E.g. log into a shell at
> Pilot or another machine and set the SSH to send a packet every 45 seconds
> or so to keep the link open.  If the connection dies, you should know it.
> Similarly, I have a server connected to Jungledisk and I get an alert in
> the logs when TWC drops out (which they do much more often than I would
> like).
>
> Some searching on this subject brought up a tool called iperf
> (https://github.com/esnet/iperf) which is supposed to test quality via
> bandwidth, delay jitter, and datagram loss.  The downside is that it
> requires a connection on two ends, though I am sure someone, including
> myself, would be willing to help you run this test, which might be good in
> that it would be fairly local to the area but still cross ISP boundaries.
>
> Last, this fairly lengthy Linux How To has a lot of suggestions for
> troubleshooting various network issues, including DNS through standard
> command line tools and mentions the diagnostic switches.  Perhaps it will
> have some suggestions that will help you collect data:
>
> http://www.linuxhomenetworking.com/wiki/index.php/Quick_HOWTO_:_Ch04_:_Simple_Network_Troubleshooting
>
> One of the big things I have discovered with dealing with ISP problems is
> that their support group likes to engage in transaction management in
> order to make you go away.  You often times need to hit them upside the
> head with actual data and evidence PROVING that THEY have a problem on
> their end. (It reminds me of when I kept getting inbound RFC 1918 packets
> FROM the TWC modem, which should never have been routed to me in the first
> place).  They denied, denied, denied until I showed them repeated logs
> with MAC addresses and proved to them that I had no devices in that IP
> range on my LAN.
>
>
> > Hi Matt-
> >
> > Can you suggest any alternatives?
> >
> > I'm looking for something that can be called from a command line, but
> also
> > lightweight enough to complete successfully under these conditions.
> >
> > I've tried the CLI version of Netalyzr (
> > http://netalyzr.icsi.berkeley.edu/index.html) but it reports so many
> > problems it's hard to pinpoint any corrective action i can take.  Plus it
> > always complains it can't measure downstream bandwidth because the
> packets
> > are blocked.
>
>


-- 
Steve Holton
sph0lt0n at gmail.com