[c-nsp] DS3 Y-Cable

[Jay Hennigan]

Peder @ NetworkOblivion wrote:
> We have a channelized DS3 into a Cisco 7505 CT3IP card.  We want a 
> backup in case something in the 7505 chassis dies, so we have an 
> identical chassis with the same cards, config and everything ready to 
> go.  The theory is that if something goes down, we just move the cables 
> and we're good to go.
> 
> A "ccie" said that we could just use a DS3 Y-cable to connect both DS3 
> cards to the one telco circuit and if one fails, it will just 
> automatically failover to the 2nd one.  That just doesn't seem right to 
> me. 

Hmmm...  I guess since they took the layer one stuff out of the lab 
things have deteriorated a bit.

You're correct, this won't work.  On the receive side (from the loop 
towards the router), it may be OK if you're lucky but you'll have a fat 
impedance bump that will very likely cause errors.  On the receive side, 
a cheapo cable-TV two-way splitter would actually do the trick nicely if 
you convert 75-ohm BNC to F.  You'll have a little over 3dB loss but 
Cisco sometimes has issues with DS3 signals that are too hot so this 
could be a blessing.

On the transmit side (from the router towards the loop) things get ugly. 
  If the cards are powered they will both be sending a bit stream, one 
idle  and the other live data.  You'll get garbage toward the line.  An 
ugly hack would be a coaxial relay powered by some type of alarm 
detector driven by SNMP or whatever.  This would be an ugly hack.

In the real world, I would recommend against it for the following reasons:

1. It is very probable that whatever cobbled-together solution you have 
to automatically "move" the live cable to the other box will have a 
shorter MTBF than the 7505 itself.  This will decrease overall 
reliability.

2.  It is also very probable that the T-3 loop and its associated 
carrier network will have a shorter MTBF than your hardware.  If this is 
a critical circuit, a second loop (with associated costs, ouch) to 
another router on the other side running VRRP is what you want.

Recommendation:  Leave the second box in the rack powered on but with 
the interfaces disconnected.  Automatically replicate the configuration 
to it from the live box with rancid, cron and TFTP, or whatever.  If you 
have a hardware failure, rapidly deploy a human to swap cables.  The 
human doesn't even need a lot of clue if the interfaces and cables are 
color coded and the boxes are labeled.  "Go into Rack 7 and move the red 
and green wires from the red and green plugs on the box labeled 
"Primary" to the red and green plugs on the box labeled "Standby.  Thanks."

> There is signaling in both directions, so how could the cable block 
> signaling one way and allow it the other way, but in the event of a 
> failure, it is smart enough to allow the other part to work.  It just 
> doesn't sound right.  Was he just plain wrong, or am I missing something?

He was just plain wrong.

> If this won't work, is there some way to provide DS3 redundancy so that 
> if one of the cards goes down, we don't have to move the cable?  That's 
> difficult if you aren't on site and don't have "hands" available.  Thanks.

Smartnet 24x7x4?  :-)

Seriously, what you are proposing is not DS3 redundancy.  It is 
interface card redundancy.  For DS3 redundancy you need a separate DS3 
to whatever is on the other end of the loop, on a different carrier 
network, with the fiber conduits a minimum of one backhoe-blade-width 
apart the whole way.

-- 
Jay Hennigan - CCIE #7880 - Network Administration - jay at west.net
NetLojix Communications, Inc.  -  http://www.netlojix.com/
WestNet:  Connecting you to the planet.  805 884-6323