[Linux-aus] strange LAN congestion problem

[Nick Edwards]

Hi group, trust we all had a great weekend.

Hope someone can help shed some light on this problem I have seen a
few times recently.
Firstly,
setup-1: 2 HP desktops, a billion router, and wireless PC
setup-2: as above with dlink switch handling the wired, passed off to
the billion
(I run internal cache DNS if that matters)

So, about one month ago, I moved my blog onto my home DSL (it's been
really reliable - more so than my old web host!), one of the pages has
a few 1MB files linked, that can often see my connection maxed out in
sending, fair enough, nothing out of place with that - unless I want
to do other things on the wired network at same time.

When blog server sends about a dozen at a time, the wired lan gets
lagged out, ssh sessions become non responsive for periods of time,
latency pinging from 0.160ms to 3000ms plus (seen on both HP's), this
is HP1 to GW and  HP1 to HP2, even HP2 to GW, and Wireless PC to
either HP, lagged, but  Wireless PC to GW no problem.

As per example of two setups, thinking the switch was screwed, I went
direct to routers inbuilt switch - still a problem, switch eliminated
, changed both HP's network cables, problem remains, my maxed upstream
of almost 40 kB/s is a sleepy yawn compared to the networks gigabit
capability, I don't see it as a PPS issue, unless both the dlink
switch (6yrs old)  and billions switch (18 months old)  use the same
identical junk components..

Now of course I can understand this lag if your on the outside, but I
can not understand why the internal network is so lagged out, and to
throw the cat amongst the pigeons, if I try simulate it from HP1
multi-send to HP2 (not going out to the net), there does not appear to
be any lag issues. I'd expect if its a buffer issue it would only
affect HP1 (blog box), not affecting HP2 in any way.

Anyways, hope someone can shed some light on this.
Nik