[30293] in North American Network Operators' Group
Re: Sonet protection usage
daemon@ATHENA.MIT.EDU (William Allen Simpson)
Wed Jul 26 07:13:28 2000
Message-ID: <397EC731.9F5118C1@greendragon.com>
Date: Wed, 26 Jul 2000 07:10:52 -0400
From: William Allen Simpson <wsimpson@greendragon.com>
MIME-Version: 1.0
To: nanog@merit.edu
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Errors-To: owner-nanog-outgoing@merit.edu
Bora Akyol wrote:
>
> Note that I am not a big fan of APS, but:
>...
> I think this is a huge oversimplification of APS and how it works. Yes APS
> allows the
> **operator** to backup various primaries to a backup. It allows 1:1 which is
> same as diverse
> circuits. This is the case of a rope that is **long enough**.
>
OK, since it "allows" 1:1, taking the same capacity as 2 circuits, why do
you think that they charge less than for 2 circuits?
Alternatively, since 2 circuits are paid for, why not use both,
doubling capacity?
My opinion (based on a fair number of years of experience) is that any
ISP foolish enough to have bought APS should also ask proof that no
other circuit is provisioned for the same APS. Once you've done that,
a class action might be in order....
APS is specified to switch a failing circuit over to a backup within 50
milliseconds. It assumes that failure is in a multiplexor. It assumes
that most of the circuits will be statistically idle. It assumes that
the individual T3 (OC-1) paths are burstable, and can be recombined at
the path or section layer. It assumes that 50 milliseconds is short
relative to switching time. In short, it assumes voice. Data doesn't
look like that at all!
If the failure is actually due to the usual circumstances, a lot of
data "circuits" fail all at once. There is no chance that they will all
be backed up.
APS (as sold) is a fraud on the uninformed.
> Well-engineered trans-oceanic links are laid such that there are at least
> two conduits running parallel
> some large distance apart.
>
And which are those? I was unaware that any were laid that way. My
information is dated on that topic, tho'. (The only one I ever viewed
was pre-optical.)
> Or you can run 1+1 IP Bonded interfaces and achieve the same effect ;-)
>
Unless it has vastly improved since I last tried it, bonding does not
work well over diverse paths, due to timing differences.
WSimpson@UMich.edu
Key fingerprint = 17 40 5E 67 15 6F 31 26 DD 0D B9 9B 6A 15 2C 32
