ARGHHH! - mix and match of 50 and 62.5 fiber

spidey07

No Lifer
Aug 4, 2000
65,469
5
76
Just looking around and trying to get some guys to test some fiber I saw we have a mix and match of 62.5 and 50 micron fiber in the data center.

WTF!!!!

There are at least 200 runs in use there and I can only imagine there is a 50 micron patch cable on a 62.5 run or riser cable.

Will it work? Maybe, maybe not...but it sure will fail spec.

oh well.
 

InlineFive

Diamond Member
Sep 20, 2003
9,599
2
0
You realize that probably the only person who is going to understand this is Jack-MDS. :)
 

spidey07

No Lifer
Aug 4, 2000
65,469
5
76
Originally posted by: PorBleemo
You realize that probably the only person who is going to understand this is Jack-MDS. :)

and scottmac and garion and nocmonkey

;)

multimode fiber optic cables come in two sizes - that's the size of the actual fiber core (the strand of glass).
 

Garion

Platinum Member
Apr 23, 2001
2,330
6
81
Yeah, that happened on our SAN, to the connection to the tape drives - It caused total and utter chaos. It would work fine, up to ~40Mb/s, then performance would TANK. Once they replaced everything with 62.5 patches (to match the premise cabling) it worked like a champ.

So, yes, it will probably seem to function, to a point. You'll see massive errors, especially when you get above a certain bitrate. Very difficult to troubleshoot, unless you follow the mantra "Always blame Layer 1, unless explicity proven otherwise"

- G
 

spidey07

No Lifer
Aug 4, 2000
65,469
5
76
well from what I gather the industry is moving to 50 micron fiber.

ScottMac?
confirm?

yes, the SAN guys are shipping 50 micron with their gear. Don't know why. SAN guys haven't got a clue.

Garion
You're right though...it will work. But how well is completely unknown.

ps - YGPM
 

spidey07

No Lifer
Aug 4, 2000
65,469
5
76
Originally posted by: Garion
Yeah, that happened on our SAN, to the connection to the tape drives - It caused total and utter chaos. It would work fine, up to ~40Mb/s, then performance would TANK. Once they replaced everything with 62.5 patches (to match the premise cabling) it worked like a champ.

So, yes, it will probably seem to function, to a point. You'll see massive errors, especially when you get above a certain bitrate. Very difficult to troubleshoot, unless you follow the mantra "Always blame Layer 1, unless explicity proven otherwise"

- G

You forgot the "spideyism"

Don't muck with the physical layer, only I didn't say muck
 

ScottMac

Moderator<br>Networking<br>Elite member
Mar 19, 2001
5,471
2
0
Fifty micron MM fiber came in about the same time as GigE.

Generally speaking, you can get more than twice the distance (550 meters versus 220 meters) on GigE / SX over 50 micron.

The distance is the same for LX/LH due to the longer wavelength (~1300 for LX versus 850 for SX) used.


I'd have to look it up, but I believe that 62.5 is still the core-of-choice for 100Meg Ethernet and that the 50 Micron stuff will reduce the recommended span length.

SO, given that the world is likely to be migrating towards GigE (and 10GigE will still be on Single-Mode, at least for a while for any kind of distance), the 50 Micron stuff is probably a safer bet overall (with single-mode being the ultimate winner in the glass arena somewhere down the road).

As long as your internal infrastructure is 220 meters or less, either will work (worse case, go for the LX and drop in some attenuation if you need to to cover the distance on 62.5).


FWIW

Scott


BTW: I spent last Saturday terminating a bunch of MMF ... what a hoot ... Used an epoxy system and ended up polishing (and polishing, and polishing, and polishing ........) : Next time, I'm going UniCam (Siecor, no polishing, ~1 minute per connection, slightly more loss, way less hand cramps).
 

nightowl

Golden Member
Oct 12, 2000
1,935
0
0
Hey, I understood all of that too. ;) Anyway, I don't know how this worked but at the place I worked at this summer a couple new guys hooked MM jumpers up to infrastructure SM fiber and it worked. I have no idea how it worked but I guess everything was lined up "just right". All I can say is that Cisco has some pretty good GBICs and even the 10G ZENPAKs too. I did not get a chance to check the interfaces but I am guessing there were some errors on it.

Scottmac, when you say the distance is the same for LX/LH are you talking about using MM fiber with that? If so, at the place where I worked (same place as above) we were able to push LX optics over 50 micron fiber to about 1km or more.
 

ScottMac

Moderator<br>Networking<br>Elite member
Mar 19, 2001
5,471
2
0
The spec, according to Cisco's site is 550 Meters (LX/LH over 50 or 62.5 micron (Which is multi-mode)).

The distance is much farther with Single Mode.


FWIW

Scott
 

n0cmonkey

Elite Member
Jun 10, 2001
42,936
1
0
Appreciate the vote of confidence Spidey07, but I haven't really worked with fiber. :eek: I figured it was cable sizes though, only thing that made sense at the time. :p
 

VirtualLarry

No Lifer
Aug 25, 2001
56,570
10,203
126
Wow, this is all somewhat interesting, although a bit beyond my personal networking experiences.

Just an idle curiousity, what type/size of fiber, do you think that Verizon is deploying, for their new upcoming "FIOS" FTTP service? Would one type make more sense than another for this purpose?

Also, how do you "coil" glass fiber-optic cable? Won't it "break"? I mean, I know that glass fiber isn't quite as brittle as normal window glass, but it still seems surprising to me that it is as flexible as it is. Are fiber patch cables, generally made out of glass or plastic or polymers?

Edit: Thanks, ScottMac, for taking the time to post a detailed explanation.
 

ScottMac

Moderator<br>Networking<br>Elite member
Mar 19, 2001
5,471
2
0
There are two basic categories of fiber: Single mode and multi-mode. The "mode" refers to how the light is propagated through the medium (usually glass).

Multi-mode relies on the "critical angle" of light ... if the light strikes a surface at a shallow enough angle, it is reflected, so transmission through multi-mode looks like "/\/\/\/\/\/\/\/\/\/." Since propagation is reliant on bouncing / reflection, the light source is generally something like an LED, usually at 650 or 850 nanometers (nm).

As mentioned above, the most common glass core sizes (for multi-mode) are 50 and 62.5 nm. To aid the internal reflective properties, a "cladding" is applied to the glass, which expands the diameter of the fiber, usually to 125 microns. The fiber diameter is usually expressed as glass core / cladding ... so standard MM fiber is usually 62.5/125 or 50/125.

Single mode fiber is much thinner (8.3 - 9 micron). Transmission is what the names says: single mode, i.e., a single straight line (no significant internal reflection) - "straight line" meaning follwing the path of the glass ... coaxial to the outside diameter .... like "--------------." The light source for SM fiber is usually some flavor of LASER / LASER diode (coherent, monochromatic light). Single mode also has cladding, usually out to 125 micron (8.3 / 125).

For either of the above, there are a bunch of "flavors." The glass comprising the fiber core is (usually) not a single consistant density. The center of the core is less dense, the outside of the core is more dense - the variable density ("stepped index" is one example) aids in reducing loss / increasing the reflective or conductive properties of the core boundries. There are a large variety of fiber types within each category ... each type is suitable for specific applications.


The glass core is so thin (in either type) and supported by coatings and insulators that it's amazingly flexible. You can tie it in a knot (not a tight knot, but a knot). Minimum bend radius for jumper-style fiber is ~1-2 inches in diameter. The downside to sharp bending is that it increases loss (in either type, but more in MMF).

The type of fiber used for the "Fiber-to-the-Premises" (FTTP) is almost certainly single mode (to accommodate the distances). It will terminate in a little box on the outside of the building, similar to a telephone "NID," and from there the individual signals will be distributed on traditional copper media.

The only exception to any of the above is the fiber used in Audio / Video (TosLink). It's a plastic fiber, somewhere around 500 micron. As fiber goes, it's total crap ... but more than sufficient for the purpose.

Also note that different applications use different frequencies (colors) of light. The A/V and 10Meg Ethernet (and other non-networking applications) use 650nm - visible Red. 100Meg Ethernet, Gig Ethernet, and other applications can use 850 nm (not visible), 1380nm (not visible), 1550nm (not visible). The point here is safety: NEVER LOOK AT A POTENTIALLY EMITTING SOURCE ... IT **WILL ** BURN THE RETINA AND PERMENANTLY DAMAGE YOUR EYES!!

That means: don't look at one end of a fiber unless you can see that the other end is not plugged into anything. Never look into a fiber optic recepticle / jack / outlet. There are "LASER goggles" available to wear if you work around an environment where LASER emitters are frequently active.

As is frequently the case, there are volumes of specific technological details for all of the above ... so the standard disclaimer applies: Some information may have been bent, folded, spindled, or mutilated for the sake of clarity and brevity.

FWIW

Scott
 

nightowl

Golden Member
Oct 12, 2000
1,935
0
0
Scott, I know Cisco says the spec for LX optics over 50u MM fiber is 550M but I was saying that we were using Cisco LX optics to span distances of 1000M or more over 50u MM fiber with no errors on the interfaces. Granted there was a limit to the distance that we finally ran into. On building was using LX optics over MM and it would randomly lose connection. The problem was that it was on the very fringe of the distance that LX optics would work over MM (the disance was over a mile I believe). Finally, after a number of lost conenctions we finally asked for the SM fiber that was recently installed but that was not turned over to us for use.
 

ScottMac

Moderator<br>Networking<br>Elite member
Mar 19, 2001
5,471
2
0
Many of the often-quoted "specifications" are actually "guidelines" designed to make it easy for someone to decide if a given system will work for them.

There are usually some "hard" specs behind the guidlines / recommendations. In the case of fiber, it is usually tied to the optical loss budget and the signal levels that the receivers can hear and tolorate..... in the case of Ethernet and other networking technologies, it is also tied to timing and delays.

Most professional organizations that install media as a business also invest in the appropriate test equipment, and understand the specifications of the media. By choosing the right media and adjunct infrastructure (patch bays, WICs, shelves, trays, connectors ...) the losses can be minimized, and the "hard spec" can be verified as "within the operational envelope" for the layer two technologies.

In your case, it sounds like you got lucky: If I was the guy in charge, and we just spent the money to run MULTIMODE fiber for ~ double the manufacturer's recommendations , and it didn't work ... the responsible person's job would be in jepordy (especially if I found out that a guidelline / specification was violated).

Optical Ethernet (10BASE-F) over MMF is usually good for two kilometers (~1.6 miles), so there is/was a fallback. To plan on running GigE so far over the recommended distance is a pure crap shoot.

I'm glad it worked for you but it doesn't say much for the planning and design capabilities of the implementers. I'd fire 'em.

Thanks for the input.

Good Luck

Scott
 

nightowl

Golden Member
Oct 12, 2000
1,935
0
0
It was just a temporary solution to move buildings over to GigE before the infrastructure SM was installed and the alternative (100M copper to fiber converters) was not a popular solution either. The MM fiber was in place from the previous generation of fiber and was used for 100M Ethernet links. If it did not work, then other alternatives were put in place. It was initially tried as something cool to do and see if it worked. Since it did work the interface was then monitored for errors to see if anything was awry. I know that as long as you are within the specs (the receiver can understand what the transmitter is sending) for the optical receivers/transmitters the whole setup will work. It also greatly depends on the quality of the fiber, the number/quality of splices, and the type of equipment that is being used. I my case the fiber path was of high quality along with the optics. It just goes to show that sometimes you have to try something and see if it works. On the other hand, long reach Ethernet does not go beyond its rating of 5000ft or at least in the situation I tried it in.