Causes of Packet Loss

[Glen Sykes]

Hi All,

I'm having some trouble at a customer site where I beleive that packet loss is being introduced through electromagnetic interference on the CAT 5 cable.

It is known that this cable runs in close proximity (around 1 foot / 300mm) to flourescent lighting, and also the length is debatable as to whether it exceeds the CAT 5 standard.

To my knowledge, the most common causes of packet loss are EMI and port speed / duplex mismatches.

Does anyone have any experience of network packet loss that doesn't fall into these categories.

For info, no firewall or router exists between the 2 devices communicating.

[Sean Lessman]

Hi All,

I'm having some trouble at a customer site where I beleive that packet loss is being introduced through electromagnetic interference on the CAT 5 cable.

It is known that this cable runs in close proximity (around 1 foot / 300mm) to flourescent lighting, and also the length is debatable as to whether it exceeds the CAT 5 standard.

To my knowledge, the most common causes of packet loss are EMI and port speed / duplex mismatches.

Does anyone have any experience of network packet loss that doesn't fall into these categories.

For info, no firewall or router exists between the 2 devices communicating.
Hi Glen,

And you are sure the right pairs are twisted in the Cat 5 cable? Just checking :)

[senthil]

Hi sean
I am having the same prob when I am trying in DSL Lines. The cat 5 twisted pairs are right, but still i am having packet loss. Is there any new changes in the twisted pairs for video ?. or the standard pairs(like stright cable).

Senthil

[Kevin]

To my knowledge, the most common causes of packet loss are EMI and port speed / duplex mismatches.

Does anyone have any experience of network packet loss that doesn't fall into these categories.


Hi Glen,

I've come across a few situations where packet loss occured and causes such as duplex mismatch were ruled out. In the end it seemed to be caused by the two network interfaces on either side being "kinda incompatible"....

Problems were fixed by puttting a simple switch of a different manufacturer or different model in the way.

Obviously all 10/100 gear should be compatible - though even though it is rare this seems not to always be the case.

Also, as you said packet loss may be being caused by the EMI and/or cable lengths being too long. In this case it is also worth trying some different and preferably modern network interfaces at either end as modern PHYs often claim to exceed the ethernet standard, both on cable length and noise rejection.

Good luck!

[Sean Lessman]

Hi sean
I am having the same prob when I am trying in DSL Lines. The cat 5 twisted pairs are right, but still i am having packet loss. Is there any new changes in the twisted pairs for video ?. or the standard pairs(like stright cable).

Senthil
No sir :-) same as any standard cat 5 IP cable. Since the purpose of twisting the cables is to cancel out noise, if you make your own cables, you need to make sure you use the wires that are twisted. I have seen plenty of times where people make their own cables and pay no attention to the color coding.

Sean

[Glen Sykes]

I highly suspect that it's the cabling. The cabling was installed by a reputable company but they did tell me that in this particular room the cable had to be run in the ceiling due a 1-piece carpet and as no cable tray was in the ceiling, the cable does run close to the flourescent lighting on occasion.

As you all know, UDP packet loss is difficult / impossible to replicate using extended Ping's etc so I'm having a hard time convincing this customer that the packet loss indicated on the endpoint is nothing other than this cabling.

Can anyone recommend a tool for testing out the capability of a cable in this way to determine 100% where the problem is? Not sure if Ethereal will do this.

[Slimey]

Fluke Networks makes some great tools to run extensive tests on just about any media type but they dont come cheap thats for sure...

[trapehzoid]

any packet loss due to EMI interference would show up as corrupted packets.. those should be seen as CRC type errors on the interfaces. If you are using managed equipment you can see these things.

Running at the edge of spec will produce things that look like packet loss.. because you are loosing data to noise and corruption. Put real equipment on the lines and watch the physical interface. Also, some equipment is better then others with dealing with this.. so sometimes putting a switch inline will help.

as others have mentioned.. there is testing gear to prove this.. but its expensive. But its something the installers should have

[Glen Sykes]

The customer is getting the installers in to prove the cables out next week now thankfully.

The cisco switch to which the VC is attached shows numerous types of error including CRC, late collisions, output errors and even a couple of interface resets.

I'm considering investing in a decent Fluke tester, but like you say it's a lot of money to spend on 'someone elses problem'. Thing is it's always my problem when network crap like this happens!

Thanks for the advice guys.

[MACC the AVCOtek]

Which Fluke are you looking at? I may have to try to convince my overseers to invest in one as well....

[Glen Sykes]

Something from the Optiview Series 2 range, they are portable and very feature rich.

Check it out here (UK site, must be a USA equivalent though).

http://www.flukenetworks.com/uk/LAN/Handheld+Testers/OptiView/Features/One+Integrated+Tool.htm