Slow Fiber?

[jward]

I had the following setup

3com 10mb switch connected to our servers, Clients
were connected to hubs via a coax to the same switch.

I replaced everything with new hardware, I now have
10/100mb switchs at both ends of a fiber run with
clients connecting to the 10/100mb switches via 10/100mb hubs

The problem I'm having is with small data transfers I have no problems, If I try to copy large files over the fiber they fail. I switched a couple users back to the coax solution and everything is ok..

This leads me to think I have a Bad fiber line. Any one have suggestions on what I can do to diagnose this. I have already put Sniffer Investigator on it only to find nothing out of the ordinary.

Thanks

 

[jimbopalmer]

One thing to watch for, (most) fiber technologies have no auto speed/auto duplex negotiation. (100baseSX is one that does)

so if you are connecting to a hub, YOU must be sure that the switch is set to half duplex, if the hub is at 10 meg, you must use a 10m meg fiber technology. (some copper to fiber converters take care of speed issues, I am not aware of one that handles duplex)

One symptom of a duplex issue on my gear is connectivity but low throughput, another is errors on the port diagnostics in the switch

I tried to remain child-like, all I acheived was childish.

 

[jimbopalmer]

Oops I first read that as fiber from switch to hub, for switch to switch just lock the ports at 100 /full.

I tried to remain child-like, all I acheived was childish.

 

[bithead9]

If you PING the server do you get ANY pack loss? I ran into a problem similar to this. And the 10MB link can be problematic. I tried dual speed HUBs/Switches and nothing worked! I had to make the connection at 10MB using a standard HUB. Sometimes the equipment will not properly negoatiate the full/half duplex. The connetion is very reliable and no packet loss

 

[abscomm]

If you have additional spare strands in your fiber cable you can swap a strand at both ends, there is no twisting or anything with fiber, just make sure that transmit at one end goes to receive at the other and vice versa.

 

[copperfiberguru]

jward,

To verify if the fiber strands in question meet expectations (if greater than 90 meters), I would suggest to measure the loss of the strands with a fiber light source and power meter. Use the below formula to determine if the loss you are reading is acceptable per the ANSI/TIA/EIA-568B.2 standard.

Cable attenuation, connector attenuation, and splice attenuation are determined by
each of the following formulas:
Cable attn. (dB) = Attn. Coefficient (dB/km) x Length (km)
Attenuation Coefficient
3.5 dB/km @ 850 nm for 62.5/125 µm or 50/125µm
1.5 dB/km @ 1300 nm for 62.5/125 µm or 50/125µm
1.0 dB/km @ 1310 nm for singlemode inside plant cable
1.0 dB/km @ 1550 nm for singlemode inside plant cable
0.5 dB/km @ 1310 nm for singlemode outside plant cable
0.5 dB/km @ 1550 nm for singlemode outside plant cable
Connector attn.(dB) = number of connector pairs x allowable connector loss
= 2 x .75 dB
= 1.5 dB
Splice Attn. (dB) = number of splices (S) x allowable splice loss (dB)
= S x 0.3 dB

*Please remember when referencing your light source and power meter to use the one jumper method.

Finding what your actual allowable loss vs. actual loss and loss from equipment (refer to manuals or contact equipment manufacturer) will give you your optical loss budget.

This will either rule out the fiber or make you look at the fiber connectors/splices for contamination.

I hope this helps.


Doug Bond, RCDD

 

[jward]

Thanks for all your input, I contacted Transition Networks because I eliminated all my other hardware and they sent out 2 replacement media converters which solved the problem.

Thanks again for your comments and suggestions.

 

[daronwilson]

I rarely use the optical loss budget calculations in troubleshooting, simply because in small installations they are overkill AND getting the data you need to do them is often tough.

Lightloss budget requires you to know:

Specific output power of the transmitter
Required receiver signal for proper decoding

Based on those two items, you get an amount of signal loss that the circuit will tolerate. Lets say you get these figures from the manufacturer, subract one from the other, and your budget is 22 dB. That means that more than 22db of attenuation/loss will cause the signal to be too low to be properly decoded. That 22 dB is your light loss budget. You can put together fiber/connectors/splices however you want so long as you don't go over 22 dB. We should consider headroom at this point. Headroom is the difference between the actual loss in the circuit (lets just say 6dB) and the total loss you could have. In this case, we would have 16dB of headroom. The circuit could degrade an additional 16dB before failure.

The next calculation is the allowable loss in your actual installed fiber and connectors as expressed above. So you go through those calculations and determine you have say 6dB in your installation. You next connect up your light source and meter and determine the actual loss on each strand, assuming the installation is of good quality, no strand should have more than 6dB total loss.

Those are good and valid tests, and the most critical one is the actual measured attenuation. In this particular example, had we gone to the trouble of calculating a light loss budget based on the hardware, we would still not know why the system wasn't working. It is complicated to measure and confirm the manufacturers specifications in the field. In this situation, it was apparently faulty hardware. The quickest troubleshooting method is probably the power source and meter.

Good Luck!


It is only my opinion, based on my experience and education...I am always willing to learn, educate me!
Daron J. Wilson, RCDD
daron.wilson@lhmorris.com