It comes from a time when wired infra was a shared medium and only one party could talk at a time. To control who talked, they passed around a token. The token would essentially take a lap around the ring before you could speak again.
Because it’s a shared medium, it’s one big collision domain.
Now, collisions are bad, mmmkay.
Modern wired infrastructure is switched. There’s some brains in the operation. The switch learns the hardware ID (unique MAC address) of every device that’s talked to it, because every frame that goes through it has the source and destination hardware ID as part of it.
As such, the switch will only forward out the port where it knows the destination is. It can only know it from one (logical) port (if there’s more than one, that’s a paddlin’). If it doesn’t know it, it’ll forward the frame out all interfaces except the one it rode in on.
Compare this with modern wireless where, aside from 802.11ax, clients just… (essentially) wait for a random amount of time, listen for a break in the signal, and take a leap of faith. It’s amazing anything works on wifi with how much modern homes stress them.
802.11ax, clients just… (essentially) wait for a random amount of time, listen for a break in the signal, and take a leap of faith.
Ethernet originally worked the same way, back when it competed directly against token ring. Ethernet won by being as reliable in real world scenarios while being cheaper to build out. Gigabit Ethernet was the first standard that insisted on full duplex only.
Half duplex mode with the collision avoidance is still actively supported for 10/100, but it is becoming very hard to find an unswitched hub. So you may have to write up your own twisted pair cables.
Yeah…kinda left a lot of the layer 1 stuff out of it when I segued into modern ethernet. I could’ve really ranted.
That was still really close to modern Ethernet when 10BaseT and 10Base2 were out. It was the switched networks and Spanning Tree that really made ethernet win out, by supporting full duplex and scaling way betterer.
Although if I remember correctly, and I could be off because most of this was before my time and I learned from a greybeard who came up on Token Ring…the same physical media for both 10Base2 (the coax with bnc ends, T’s at each station, and a terminator on each end) and later 10BaseT (UTP CAT 3, if we are being contemporary) Ethernet did end up getting used for Token Ring as well, just different hardware. And I think IBM was actually able to squeeze more bandwidth out of the same wires for a while, too.
Honestly if not for switched Ethernet and Spanning Tree, Token Ring would’ve had several more years of life left in it, at least.
Modern 802.11ax (Wifi6) borrows a little bit from the methods of token ring by having the AP essentially schedule timeslots for each of the clients…but it won’t make much improvement till most (if not all) of the segment (that is, devices connected to any one radio) is ax or higher. I believe (and don’t quote me on this), that is very similar in concept to how the later generation token ring MAUs worked.
You can still do 10Half on some high end managed switches. But most SFPs and mGig ports won’t go down that low, and mid-range enterprise access switches are getting fewer and fewer 10/100/1000 ports.
Good. Maybe that means they’ll finally upgrade the damn badge readers.
If it’s stupid and it works, it ain’t stupid.
It comes from a time when wired infra was a shared medium and only one party could talk at a time. To control who talked, they passed around a token. The token would essentially take a lap around the ring before you could speak again.
Because it’s a shared medium, it’s one big collision domain.
Now, collisions are bad, mmmkay.
Modern wired infrastructure is switched. There’s some brains in the operation. The switch learns the hardware ID (unique MAC address) of every device that’s talked to it, because every frame that goes through it has the source and destination hardware ID as part of it.
As such, the switch will only forward out the port where it knows the destination is. It can only know it from one (logical) port (if there’s more than one, that’s a paddlin’). If it doesn’t know it, it’ll forward the frame out all interfaces except the one it rode in on.
Compare this with modern wireless where, aside from 802.11ax, clients just… (essentially) wait for a random amount of time, listen for a break in the signal, and take a leap of faith. It’s amazing anything works on wifi with how much modern homes stress them.
Ethernet originally worked the same way, back when it competed directly against token ring. Ethernet won by being as reliable in real world scenarios while being cheaper to build out. Gigabit Ethernet was the first standard that insisted on full duplex only.
Half duplex mode with the collision avoidance is still actively supported for 10/100, but it is becoming very hard to find an unswitched hub. So you may have to write up your own twisted pair cables.
Yeah…kinda left a lot of the layer 1 stuff out of it when I segued into modern ethernet. I could’ve really ranted.
That was still really close to modern Ethernet when 10BaseT and 10Base2 were out. It was the switched networks and Spanning Tree that really made ethernet win out, by supporting full duplex and scaling way betterer.
Although if I remember correctly, and I could be off because most of this was before my time and I learned from a greybeard who came up on Token Ring…the same physical media for both 10Base2 (the coax with bnc ends, T’s at each station, and a terminator on each end) and later 10BaseT (UTP CAT 3, if we are being contemporary) Ethernet did end up getting used for Token Ring as well, just different hardware. And I think IBM was actually able to squeeze more bandwidth out of the same wires for a while, too.
Honestly if not for switched Ethernet and Spanning Tree, Token Ring would’ve had several more years of life left in it, at least.
Modern 802.11ax (Wifi6) borrows a little bit from the methods of token ring by having the AP essentially schedule timeslots for each of the clients…but it won’t make much improvement till most (if not all) of the segment (that is, devices connected to any one radio) is ax or higher. I believe (and don’t quote me on this), that is very similar in concept to how the later generation token ring MAUs worked.
You can still do 10Half on some high end managed switches. But most SFPs and mGig ports won’t go down that low, and mid-range enterprise access switches are getting fewer and fewer 10/100/1000 ports.
Good. Maybe that means they’ll finally upgrade the damn badge readers.