CRC Errors

No segments should have more than one CRC error per million bytes of data, or 0.0001 percent CRC errors on each segment.

This can be represented by the following formula:

NOTE

The term CRCs is often used for brevity when discussing CRC errors.

In the following code, the total amount of data is 742089835 bytes [Input Bytes (305539992) + Output Bytes (436549843) = Total Bytes (742089835)]. The total number of CRCs indicated is 34215.

2295197 packets input, 305539992 bytes, 0 no buffer Received 1925500 broadcasts, 0 runts, 0 giants 3 input errors, 34215 CRC, 0 frame, 0 overrun, 0 ignored, 0 input packets with dribble condition detected 3594664 packets output, 436549843 bytes, 0 underruns

0 abort

Using the previous formula, (CRCs / total bytes) x 100, you can determine the following:

This is an unacceptable amount of CRC errors because the total here (0.0046 percent) is greater than 0.0001 percent of all CRCs. An acceptable threshold would be 742 CRC errors (742,089,835 x 0.0001 percent = 742.089).

A case in which the number of CRC errors is high, but the number of collisions is not proportionately high, is usually an indication of excessive noise. In this case, the following actions should be taken:

1. Check the cables to determine whether any are damaged.

2. Look for badly spliced taps, causing reflections.

3. If using 100BaseTX, make sure Category 5 cabling is being used and not another type, such as Category 3. Ethernet Collisions

On Ethernet segments, less than 0.1 percent of the frames that are identified as packets in the show interface output are collisions.

This can be represented by the following formula:

(Collisions / (Input packets + Output packets)) x 100 < 0.1 %

The following code explains the output from Ethernet segments from the show interface command:

2295197 packets input, 305539992 bytes, 0 no buffer Received 1925500 broadcasts, 0 runts, 0 giants

3 input errors, 34215 CRC, 0 frame, 0 overrun, 0 ignored, 0 abort 0 input packets with dribble condition detected 3594664 packets output, 436549843 bytes, 0 underruns 8 output errors, 1790 collisions, 10 interface resets, 0 restarts 0 output buffer failures, 0 output buffers swapped out

Using the previous formula, (Collisions / (Input packets + Output packets)) x 100, you can determine the following:

This is an acceptable amount of collision errors because this is less than 0.1 percent of collisions. An acceptable maximum threshold would be 5890 collisions (2295197 + 3594664 x 0.1 percent = 5889.861).

Collisions are part of normal operation for Ethernet and IEEE 802.3 (CSMA/CD) networks. Excessive collisions could be caused by a number of factors. If an excessive number of collisions is detected on the interface, the following steps should be taken to correct them:

1. Use a TDR to find any unterminated Ethernet cables.

2. Look for a jabbering transceiver attached to a host. (This might require host-by-host inspection or the use of a protocol analyzer.)

3. Use a class 1 cable scanner to make sure cabling certifies from end to end. (Category 3, 5, and so on are end-to-end specs, and a poor punch-down can cause cabling systems to be out of spec.)

A network analyzer is necessary to determine the type of collision that is detected on the network segment. Three types of collisions can usually be found on an Ethernet segment:

• Local collisions

• Remote collisions

• Late collisions Local Collisions

On a coaxial cable segment (10Base2 and 10Base5), the signal traverses the cable until it encounters the signal from the other host. The signals then overlap, canceling some parts of the signal out and reinforcing (doubling) other parts. The "doubling" of the signal pushes the voltage level of the signal beyond the maximum-allowed transmit voltage level. All of the hosts then sense this over-voltage condition on the cable segment as a local collision. On unshielded twisted-pair (UTP) cable (10Base-T), a local collision is detected only when a host detects a signal on the receive (RX) pair at the same time that it is transmitting on the transmit (TX) pair.

Although local collisions are part of normal operation on a CSMA/CD network, they could be caused by a number of factors:

• Overloaded network segment

• Faulty or marginal network interface card (NIC)

• Ethernet transceiver fault

• Ethernet repeater fault

• Illegal hardware configuration

• Ethernet cable fault

• Bad or poor host termination

• Bad grounding

• Induced noise on the segment (improperly shielded cabling)

Was this article helpful?

0 0

Post a comment