Network Debugging

[Hornet]

On 7/7/05, Mark Bucciarelli <mark at gaiahost.coop> wrote:
> I'm trying to improve the performance of our rural homebrew wireless
> broadband and am hoping some of the folks here can give me a pointer or
> two as to what network monitoring tools I should use.
> 
> Background:
> 
> - my internet access is five wireless and five wired hops downstream
>   from a 1/2 T1 (ten Buffalo AirStation G54 routers in all).
> 
> - my connection has a lot of jitter--ping's usually vary from 10ms to
>   150ms within a two second window
> 
> - FWIU, jitter is related to congestion
> 
> - I have setup a FreeBSD box to monitor [1] each router along the path
>   using smokeping.
> 
> The smokeping charts are showing me some interesting stuff.  Here is
> some data from the past three hours (I am using the smokeping default of
> 20 pings sent every five minutes):
> 
>    |--------- Building 1 --------------|              |--- Bldg 0 --...
> 
>    +-----------+          +------------+              +-----------+
>    | .203      |          | .202       |              | .201      |
>    | Router In |<- wire ->| Router Out |<- wireless ->| Router In |
>    +-----------+          +------------+              +-----------+
> 
> avg RTT: 7.3ms              12.1ms                      7.8ms
> 
>  % lost: 2.37%              14.25%                      2.64%
> 
> max RTT: 20ms               80ms                        13ms
> 
> 
> My FreeBSD box is a four more wireless hops to the left of .203.
> 
> A slew of questions ...
> 
> What is going on here?
> 
> I am confused by the max RTT readings and packet loss stats for .202 and
> .201.  How can a router further away from me have better performance?

Most routers put a lower priority on ICMP, If the middle router  has a
higher load on it (which by your diagram it should), then the B router
would be slower to respond to ICMP.

Try using MTR (Matt's traceroute) It will give your real time stats on
your network hack.

> 
> Over the past 13 hours, the averages are consistent with the three-hour
> averages, while the Max RTT discrepancies are even higher:
> 
>       .203 / .202 / .201 = 20ms / 145ms / 13ms.
> 
> Is .202 congested?
> 
> Is the .202 router "bad"?
> 
> How can I debug this further?  SNMP?
> 
> If SNMP, what values should I track/inspect?
>   - # of packets with errors?
>   - # of queued packets?
>   - ??
> 
> Thanks for any pointers,
> 
> m
> 
> [1] Pentium II 350MHz with 4 Gb drive, underclocked to 100MHz so I can
>     turn off the power supply fan and make it real quiet.  :)
> 
> _______________________________________________
> freebsd-questions at freebsd.org mailing list
> http://lists.freebsd.org/mailman/listinfo/freebsd-questions
> To unsubscribe, send any mail to "freebsd-questions-unsubscribe at freebsd.org"
>