1

Problem: any sleeping computer on a gigabit network causes all multicast traffic to drop to 10Mbps.

Solution: provide customers with managed gigabit switches and train them to configure the switches manually. Not an ideal solution.

I discovered today another workaround… the cause is 802.3x flow-control. Disabling flow-control fixes the problem:

sudo ifconfig en0 -mediaopt flow-control

However this means that every device on every customers network would need to turn off auto negotiation, then manually turn flow-control off. Turning off auto-negotiation is a very bad idea. Not a workable solution.

One example: I notice that on my dumb, old gigabit switch,

  • sleeping machine has flow control ON , link light ON, 1000Mbps light is off
  • sleeping machine has flow control OFF, link light ON, 1000Mbps light is lit.

My question is: Is there any RFC/protocol which would allow traffic to ignore flow-control. (Probably not)

Is there any RFC/protocol/hack to discover whether a sleeping device on a network uses flow-control… (Maybe?)

Is there any RFC/protocol/hack to discover what is the auto-negotiated speed of each device on its immediate switch (albeit dumb switch)?

This would at least allow my software to warn a user why their multicast performance is 10Mbps.

  • Please do not cross-post on multiple SE sites. You should pick one site, and delete the question on the other sites. – Ron Maupin Oct 3 '16 at 20:02
  • Did any answer help you? If so, you should accept the answer so that the question doesn't keep popping up forever, looking for an answer. Alternatively, you could provide and accept your own answer. – Ron Maupin Aug 15 '17 at 4:11
1

My question is: Is there any RFC/protocol which would allow traffic to ignore flow-control. (Probably not)

Ethernet is not controlled by the IETF RFCs, it is from the IEEE 802.3 committee. Many devices do ignore ethernet flow control; it is poorly supported, and it is not required. Even among network devices which do support it, you must often enable it.

Is there any RFC/protocol/hack to discover whether a sleeping device on a network uses flow-control… (Maybe?)

Sleeping devices don't tell you anything; you must first wake them. Hosts are off-topic here. There will be no RFC because the IETF doesn't deal with ethernet standards, and I know of no such protocol from the IEEE.

Is there any RFC/protocol/hack to discover what is the auto-negotiated speed of each device on its immediate switch (albeit dumb switch)?

If a switch or the host supports SNMP, then you could use SNMP to discover such things, but that is unlikely for a dumb switch.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.