Ethernet Comms Confusion

[daba]

@krakenfan69

Switch versus Hub - simply put.....

EtherNet works using CSMACD technology - that's Carrier Sense, Multiple Access, Collision Detect.

Carrier Sense - nodes on the network listen for no "carrier" signal, which is when no nodes are transmitting.

Because of Multiple Access, all nodes have equal priority, and can start transmitting simultaneously.

This will inevitably lead to "collisions", which all transmitting nodes will detect, and they will all stop transmitting.
​

Then each node waits a random time delay, before waiting for a quiet network again. So one of the nodes will succeed, gaining the network resource for its sole use.

To be efficient, the network components should reduce the number of collisions, ideally to zero.

A Hub is just a dumb "port expander", meaning that a data-packet sent through one port will be relayed out of all the other hub ports. This of course escalates the possibility of collisions.

A Switch, however, will learn what IP addresses can be reached through each port. It will read the destination address embedded in each data-packet, and route it out to the port that is connected to the destination. This dramatically reduces the "collisions". If the switch ports only have one node attached to them, then collisions are eradicated completely (assuming that the ports are working in full-duplex mode). A switch will also allow devices at differen speeds (e.g. 10MB/s & 100MB/s) to communicate with each other.

Don't even consider using a hub nowadays.

 

[The Plc Kid]

DHCP persistence - Not quite the same as vanilla DHCP, which just gives a spare address to the first device that asks for one. And gotta be a PITA if the switch fails.

If the switch fails you are SOL anyway. If you have a spare you just swap the flash card. bubba can do it.

I prefer, and recommend, to fix IP addresses on automation nets.

I agree with you 100% and you can have both. Set them static but set the same address in the DHCP pool and assign to that specific port. label you cables and on your prints to reflect this. This way you have the rock solid reliability of static addresses but if you lose a drive at 2:00 AM bubba can put one in turn the system on and the switch will assign a address and if it's a rockwell drive and contrologix PLC of the correct firmware auto device replacement will load the drives program and you are back in production. I use this method a lot.

 

[The Plc Kid]

Also you can do DHCP per port with almost any managed switch worth it's salt. Sometimes the terminolgy for that function is a little different though.