Super User is a question and answer site for computer enthusiasts and power users. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I am reading the wikipedia article of mac addresses and it states that the least significant bit of the most significant byte of the mac address of a NIC indicates the following:

If the least significant bit of the most significant octet of an address is set to 0 (zero), the frame is meant to reach only one receiving NIC.[5] This type of transmission is called unicast. A unicast frame is transmitted to all nodes within the collision domain, which typically ends at the nearest network switch or router.

I have two computers each has a mac address where the least significant bit of the most significant byte is 0. That means when a frame is sent from the NIC over the ethernet line it is intended to be sent to one receiving NIC. Now I have both my computers wired to a Belkin router. A router avoids collision domain because each NIC has its own dedicated ethernet line. So what kind of collision domain is this article talking about?

share|improve this question
up vote 2 down vote accepted

"Collision domain" is sometimes used to mean what should more correctly these days be "broadcast domain" - a set of nodes interconnected in some manner whereby any node may send an Ethernet frame to any other node, and broadcast frames propagate to all nodes interconnected.

@Frank Thomas's comment below is very true, though.

Now I have both my computers wired to a Belkin router.

If your Belkin router is a typical consumer device with an Internet or WAN ethernet port, and a set of four LAN ethernet ports. This device is actually a combination of a router plus a switch.

Pure routers only have two ethernet ports, sometimes labeled WAN and LAN. You are then responsible for connecting a switch to the LAN port if you want to connect multiple systems to it.

With your Belkin router, the switch part of the device is forwarding LAN-originating broadcast Ethernet frames. Unless the switch part of the device is implemented in software, this doesn't even pass through the routing software on device. Logically, the setup is like a pure router + switch on the LAN port above.

Switches keep a port to MAC mapping in what is typically called CAM tables.

Routers, computers, and other IP capable Ethernet devices keep an IP to MAC mapping called the ARP table.

share|improve this answer
If a NIC on one computer sends ethernet data destined to another computer, since both machines are connected to the router, the frame first goes through the router, and then the router keeps some kind of mapping of mac addresses, so it knows which mac address to send to. THat's how I thought it worked. Now you are saying that the router sends a broadcast to all machines connected to it to find out which computer to send the packet to? – JohnMerlino Mar 31 '14 at 3:00
Yes, the language is quite badly used in the area. – vonbrand Mar 31 '14 at 3:01
@JohnMerlino, the router keeps a table (ARP table) mapping IP addresses to MAC addresses, and uses this to direct traffic. – vonbrand Mar 31 '14 at 3:02
switches propagate L2 broadcasts (like DHCP and ARP requests), but have micro-segmented collision domains, so the two terms are not entirely synonymous. you use switches to break up collision domains, and VLANS/Routers to break up broadcast domains. – Frank Thomas Mar 31 '14 at 3:06

Those "unicast" addresses are actually the normal Ethernet addresses. Other alternatives are (rarely used) multicast addresses and (somewhat common) the broadcast address. NICs can be configured to read frames addressed to several addresses (either multicast or unicast), and always read frames addressed to the broadcast address.

In the original Ethernet, all NICs were connected to a single cable, and thus all could interfere with each other (causing collisions between their frames, thus "collission domain"). Later bridges were introduced, which split one cable into two pieces, and only forward frames from one to the other when the bridge knows the recipient is on the other branch. This creates separate collision domains. A hub essentially folds the long cable into cables to and fro the machines, nothing really changes. A switch is essentially a multi-leg bridge, with one leg going to each machine. The machine you have (and what you use locally) is a switch, one of whose legs is connected to a router, a machine that connects two IP networks (one of them "outside").

share|improve this answer

A collision-domain refers to a length of shared network cable in a bus topology, like the old ethernet networks that used a single COAX cable to connect to every host on the segment. This is a very very old design, and you will probably never see one in reality. common hardware included the bus cable, "vampire taps" (wires that physically peirce the bus for contact) which connect the hosts to the bus, bridges for connecting two busses together, and hubs (bridges that can connect multiple segments).

They are called collision domains, because only one computer can send at one time. if two attempt to send, there will be a collision, and the data that both hosts were trying to send is corrupted.

on these kinds of networks, a host listens for anyone sending, and then attempts to send a signal. it then listens for sign that a collision occurred. if so, it pauses for a bit and then tries again.

a bus topology looks kinda like (X is the end of the bus, H is a host,| is a drop line attached by a vampire tap, and = is the coax bus).

       H                   H
       |                   |
                  |                 |
                  H                 H
share|improve this answer

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .