Ethernet over mains power experience

[timcoote]

Hullo
I'm streaming video over an Ethernet over mains link. Mostly the protocols work ok. However, I'm seeing periodic changes in ping rtts moving from ~9ms to ~100ms, which is enough to cause the stream to stutter annoyingly.

so my question is: is the variation in rtt a fundamental problem with Ethernet over mains, or is it more likely to be a problem with either the kit that I've got or some artefact of the wiring in my house?

tia
Tim

[jpollard]

Possibly neither.

That amount of change is almost like a route switching around a failing router.

If it does it continuously (say every 10-15 seconds) it can also be caused by a router dropping a route and having to rediscover it.

[timcoote]

the frequency varies significantly between hardly ever to every 2-3 seconds, even over a single point to point link with a linux box at each end (which nevertheless does have a router on one of the boxes at one end of the link to get to the rest of the world).

How do I diagnose such a type of failure ? run up wireshark and look for relevant packet discovery packet types?

[I do wish that network protocols followed good engineering practice and 'failed fast', rather than struggling on]

[jpollard]

Try a ping sequence to the system you are targeting and see if it also gets the breaks.

Another test is traceroute to the router and target host. If the break occurs during the trace you should see a sudden jump in the timing between two hosts. That break should identify where (between which two) hosts involved.

Network protocols do fail fast - when it is a failure. Hosts on either side of a failure do have the capability to try and route around a failing node. The next level problem is "how long to wait before it is a failure"...

A rate of 2-3 seconds also can be caused by a bad/corroded connection. Sometimes just unplugging and plugging back in can clean the contacts enough. This occurs more often in a damp environment or wintertime under cooler periods.

[timcoote]

I'm detecting the variation using ping. I did quite an extended experiment over a number of days unplugging all of the devices on the mains, which eventually removed the problem, but made the house dark, cold and quiet :-) Small 'wart' chargers were a particular cause of noise (eg wii games, phone chargers).

However, I'd have thought that such noise should be filtered out - and this is really what I'm trying to establish: is the RTT variation that I'm seeing typical of ethernet over mains, or unusual.

Traceroute does not show me much as I'm not sure that the protocol stacks on the half bridges are responding at the same rate as the through traffic. (this seems to be the common behaviour of low cost devices).

[failfast includes logging the failure - I'll drop this tho' as I don't think it's a useful discussion point]

I don't think that there's a specific connection problem, but there could be - it is an oldish house - but I'd expect a more regular pattern of noise for a poor connection.