I have a situation where I want to run Nx24 pairs of GE across a datacenter to several different customers. Runs are about 200meters max.
When running say 24-pairs of multi-mode across a datacenter, I have considered a few solutions, but am not sure what is common/best practice.
a) Find/adapt a 24/48 thread inside-plant cable (either multimode, or condition single mode) and connectorize the ends. Adv: Clean, Single, high density cable runs, Dis: Not sure if such a beast exists in multimode, and the whole cable has to be replaced/made redundant if one fiber dies and you need a critical restore, may need a break out shelf.
b) Run 24 duplex MM cables of the proper lengths. Adv: Easy to trace, color code, understand. Easy to replace/repair one cable should something untoward occur. Can buy/stock pre-terminated cables of the proper length for easy restore. Dis: Lots of cables, more riser space.
c) ??
Look into MPO cabling
MPO uses fiber ribbon cables the most common of which is 6x2
six strands by two layers
Panduit has several solutions which use cartridges so you get a
cartridge with your desired termination type and run the MPO cable between
the cartridges.
This cabling under another name is also used for IBM Mainframe channel
connections
Scott C. McGrath
I assume multiplexing up to 10Gb (possibly two links thereof) and then
back down is cost-prohibitive? That's probably the "best" practice.
Thor
Thor Lancelot Simon wrote:
>I assume multiplexing up to 10Gb (possibly two links thereof) and then
>back down is cost-prohibitive? That's probably the "best" practice.
It's best practice to put two new points of failure (mux + demux) in a
200m fiber run?
Well, that depends. To begin with, it's not one run, it's 24 runs.
Deepak described the cost of those 24 runs as:
I priced up one of these runs at 100m, and I was seeing a list price in
the ballpark of $2500-$3000 plenum. So I figured it was worth asking if
here is a better way when we're talking about N times that number. 
So, to take his lower estimate 24 x $2500, we're talking about $60,000
worth of cable -- and all the bulk and management hassle of 48 strands
of fibre for what is in one sense logically a single run.
It still probably doesn't cover the cost of muxing it up and back down,
but particularly when you consider that space for 48 strands isn't free
either, it is certainly worth thinking about.
I was a little surprised by the $2500/pair figure but that's what he
said.
Thor
a) Find/adapt a 24/48 thread inside-plant cable (either
multimode, or condition single mode) and connectorize the
ends. Adv: Clean, Single, high density cable runs, Dis: Not
sure if such a beast exists in multimode, and the whole cable
has to be replaced/made redundant if one fiber dies and you
need a critical restore, may need a break out shelf.
We use multicore fibre cables in our datacentre and nodes.
In Europe we can obtain these to order on specific
size. If we need to bring these back to a central area we
use an ODF and and then patch accordingly with fibre boxes
in cabinets. The multicores are in a rugged PVC type plastic
sheath [same type of plastic/pvc that is used for gas piping
in the streets here in Europe]. You have to do some serious
hacking to damage this type of cable, and if you've ever
been to Telehouse in London you'll know how the type
of situation that I mean.
This is the company we use in the UK to do a lot
of this work
http://www.mainframecomms.co.uk/products_cables.html
I can't recommend them highly enough.
You may also want to look at passive optical stuff that you
can use to cheaply use to do cdwm.
Regards,
Neil.