Ethernet Layer 1 – Reasons for Excessive Collisions in Hub Networks

ethernetlayer1

I ran a test-setup which caused excessive amount of collisions on Cisco FastHub 400 port. This caused switch to isolate this port. For example in case of port 6, this has happened 766 times:

Hub#sh interfaces fastethernet 1/0/6
FastEthernet 1/0/6 is link-up
Hardware is built-in 100Base-Tx
Description: 
Port admin: enabled
Port partition: not Auto Partitioned
Port operation: Operational
Speed admin: Auto Negotiation
Speed status: 100M
Last Source MAC Address: 00-1D-09-F0-92-AC

Counters
--------------------------------------------------------------------
Bytes Received         1343247196  Frame Received            3719902
CRC Errors                      0  Alignment Errors                0
Frame Too Long                  0  Short Events                    0
Port Runts                      0  Collisions                1145745
Late Events                     1  Data Rate Mismatches            0
Auto Partitions               766  Total Errors                    1

Hub#

What was the main reason in hub networks to cause such high amount of collisions that hub had to isolate the port from rest of the network?

Best Answer

Ethernet originally utilized CSMA/CD (carrier sense multiple access with collision detection). Note that "collision" is part of the name. CSMA/CD is required in a hubbed environment due to the nature of how a hub operates.

Hubs operate at the physical layer by repeating a signal received on one port out all the other ports on the device. If a hub receives a signal on two different ports at the same time, the resulting "merged" signal that gets repeated is meaningless.

Ultimately all this means is that collisions on a hub are to be entirely expected. However in your case, the number of collisions is very high, over 30% of your traffic. So what can create a situation like you are facing? Here is a non-exhaustive list:

  1. Duplex mismatch with a connected device - the connected device would have to be configured for full duplex or auto-negotiation is failing.
  2. A loop on the network.
  3. Many devices on the collision domain; the more devices you have the greater the chance of collisions occurring.
  4. Network design that violates the 5-4-3 rule.
  5. A device that is defective/malfunctioning.
  6. A newer device that does not behave correctly in a half-duplex environment, sometimes related to a driver issue.
Related Topic