Right off the back, do switches experience collisions. I am asking because, if my understanding is correct, each switch port is it's own collision domain. Furthermore, with filtering capabilities, the switch will filter traffic when source and destination MACs are on the same port.
The second thing baffling is that, if switches experience little to no collisions (an assumption), what is the point of using Fragment Free mode?
Explanations will be greatly appreciated. Thanks in advance.
A collision domain is any part of a network where collisions can occur.
Switches create micro segmentation. What happens ins a switch will create a virtual circuit between each port on the inside. Now in theory each switch port is it's own collision domain, however inside that collision domain there can be a few factors that will still allow collisions to occur. First off, if the port is in half duplex (one way at a time) a collision can occur. With Full-duplex collisions won't occur because there is a path for both directions. Next factor is what is connected to it. If it is a device such as another switch, a router, a computer, ect then the switch port and the device are in a separate collision domain from the rest of the network. If a hub or repeater are attached to that switch port and attached to that hub are 10 devices, then the 10 devices and the switch port are in the same collision domain.
Correct if the destination MAC is listed as the same port as the incoming port the switch will filter the frame.
If 0007.ec91.0e01 sends a packet to 0007.ec91.0e02 the switch will receive the frame reference the destination to its table then drop it.
Mac Address Table
Vlan Mac Address Type Ports
1 0007.ec91.0e01 DYNAMIC Fa0/1
1 0007.ec91.0e02 DYNAMIC Fa0/1
Fragment-free switching uses the theory that if an error is going to occur in a packet it will happen in the first 64bits of the packet. If those bits are wrong then the packet is dropped. Line interference or collisions can ruin those first 64bits. So not only does it protect against collisions it also protects against line interference.
Hope this helps, if you have any other questions feel free to ask. And if I'm wrong about something please let me know
Kindly clarify with this scenario.
6 workstations (WkSt1, WkSt2, WkSt3,...) are connected to a switch named SW1 on ports 1 through 6 respectively. A couple of server units (Svr1 and Svr2) are connected on switch SW1's ports 7 and 8. An eight port hub fully populated with PCs is connected to switch SW1's on port 12.
Note that all switch ports are in full duplex mode.
Mac Address Table
1 0007.ec91.0e51 DYNAMIC Fa0/1
1 0007.ec91.0e52 DYNAMIC Fa0/2
1 0007.ec91.0e53 DYNAMIC Fa0/3
1 0007.ec91.0e54 DYNAMIC Fa0/4
1 0007.ec91.0e55 DYNAMIC Fa0/5
1 0007.ec91.0e56 DYNAMIC Fa0/6
1 0007.ec91.0ea1 DYNAMIC Fa0/7
1 0007.ec91.0ea2 DYNAMIC Fa0/8
1 0007.ec91.0e01 DYNAMIC Fa0/12
1 0007.ec91.0e02 DYNAMIC Fa0/12
1 0007.ec91.0e03 DYNAMIC Fa0/12
1 0007.ec91.0e04 DYNAMIC Fa0/12
1 0007.ec91.0e05 DYNAMIC Fa0/12
1 0007.ec91.0e06 DYNAMIC Fa0/12
1 0007.ec91.0e07 DYNAMIC Fa0/12
1 0007.ec91.0e08 DYNAMIC Fa0/12
Will a collision triggered on PCs connected to the hub in the scenario show up (propagate) to port 1 through 8 on switch SW1?
If the answer is no (my assumption) what will be the benefit to turn on Fragment Free switching on switch SW1?
So here is what your LAN looks like.
With switches either you manualy set the duplex or its auto negotiated. When a cisco switch is attached to a hub it will set itself to half-duplex, and setting it to full will give you errors ( I believe).
Each circle represents a collision domain. Switches break up collision domains, where hubs and repeaters extend them. Each device attached to the hub shares the same line and is one collision domain. The rest of the network can run on full-duplex but the hubbed section can't. Therefore CSMA/CD must be run.
The mac table on the switch would look like what you said.
From my understanding a collision only affects that domain and won't affect the rest of the network. But even still, errors can occur and still lead to problems. With cut through switching the frames are sent error or not. This can put burden on the backbone if the frame travels several switches gets to its destination and is dropped because its bad. Fragment-free allows for the reduction of wasteful frame usage. (Proper terminology not used). The type of switching depends on the size and needs of the network. Also I do not think it is possible to have port based switching types but I could be wrong.
So in short to answer your question, it only affects the computers in the collision domain (to my knowledge) and fragment-free switching can be beneficial to the entire network as to prevent corrupt frames form taking up bandwidth.
If you have any other questions feel free to let me know. I hope this helped.
Bottom line is you will never get a collision on a switch port, since there is no other device on the port for the data to collide with. You will most likely always get a collision on a hub if you have 2 or more nodes attached.
Nice job guys! This is a very informative post!
In this topology if there is a collision at the hub it would reach the switch port as this is one single collision domain. At least that has always been my understanding of it. I think this is what you were saying but wanted to clarify just in case someone read and thought that a switch could never see a collision. Also, in this scenario the switch port leading to the hub would have to be half-duplex and CSMA/CD would have to be used. I think that the best way to word that would be a switchport in full-duplex can never have a collision!
Sorry, had to edit post.. Typed CDMA/CD! LoL