[mrtg] Re: Live vs Gold Pipes

[Calvert, Neil]

might as well show off some peripheral knowledge, eh? 

I'm ASSUMING here you're talking about Cabletron Spectrum. If you're not,
then ignore me.

1)	Live pipes are a means of polling the two ports that make up a link.
They essentially instruct an inference handler to keep track of this link by
monitoring the ifoperstatus and ifadminstatus of the ports.
2)	If you have a live pipe and a link goes down, then you should
receive an additional alarm. Also, the red line between devices is quite
obvious to an operator that something has occured. If live pipes are not
enabled, the link will remain gold (resolved).
3 & 4)	The 'need' to have live pipes is really dependent on your network.
They do create additional polling traffic and additional work for the
server. As for backplane links, you can model this as a live link, but it
would probably be overkill. If you have a problem with backplane traffic it
would probably be a full hardware failure in which case the router/switch
blade itself should be alarming.

You do get fault isolation w/out live pipes. Fault isolation is accomplished
by having fully resolved (port to port) links. This gives Spectrum the
intelligence needed to say 'Ok, this port went down, I know this WAN network
is off this port, so I'll make that Grey, since it's now unreachable'. Live
pipes just give you more alarms, and also more visual cues that something is
awry.

Hope this helps.

Neil


-----Original Message-----
From: Bradford Woodcock [mailto:bwoodcoc at Genuity.com]
Sent: Thursday, June 22, 2000 11:13 AM
To: mrtg at list.ee.ethz.ch
Subject: [mrtg] Live vs Gold Pipes



Greetings,

I am hoping someone on the list can answer these questions.

1. What *exactly* do live pipes do?
2. Is the amount of extra work for live pipes significant?
3. Do I need pipes to be live even for backplane links?
4. If I leave the pipes gold will I still get up/down notification with
full fault isolation? 

I ask these questions because my company is in a transition state. And
because of this I will be forced to heap a LOT of switch modules onto an
Ultra 10 server (~350) and I'm not sure it will handle it. I'm hoping if I
don't turn on live pipes the server may fair better. I have already turned
off switch interface logging. I of course still need basic up/down
monitoring with fault isolation capabilities. Thanks in advance for your
experiences.

cheers,
brad...