From: Scott Vermillion (scott_ccie_list@it-ag.com)
Date: Sat Nov 10 2007 - 01:48:07 ART
Hi Carlos,
I used to work in the WAN world. Commercial circuits are generally sold
with a service level agreement of some sort. So I might want to buy a T-3
from points A to B. It will cost me more to get a guaranteed availability
of 99.9997% (what we referred to as "five nines and a seven") vs. simply
99.97 (three nines and a seven). What is meant by "availability" generally
is that data is being delivered across the circuit *and* that a particular
Bit Error Rate threshold is not being exceeded (so you could be getting
999,999 unerrored bits for every errored bit, and you could still regard the
circuit to be unavailable).
Of course, you do not necessarily lose that .0003 or .03 percent of traffic,
nor do you necessarily ever exceed your BER threshold for that percentage of
time; it's simply the case that you cannot penalize the carrier for circuit
unavailability up to that amount of time (per month or per year, whatever).
Having said that, you generally need to plan for that level of loss whether
it's realized or not. So I might talk to the application developers and try
to figure out what kind of packet loss they can experience before things go
to pot. I might talk to the business minded folk and determine at what
point not having service active is more expensive than a really tight SLA.
Then I base my contract around that. This becomes far more critical for
real-time apps than for others. If I was putting a commercial web server
up, there's no way I would pay for a five nines circuit (this amounts to
something like only five minutes of downtime or BER exceeded per year)!
However, if I was going to be pushing a real-time video feed over the
circuit or something, I'd be more inclined to build in some tighter
parameters with my carrier(s).
Bottom line is that no circuit is perfect, but the rated availability is a
worst-case thing and it's not necessarily so that you will actually
experience that level of loss/unavailability. Just plan to...
Regards,
Scott
-----Original Message-----
From: nobody@groupstudy.com [mailto:nobody@groupstudy.com] On Behalf Of
Carlos glez
Sent: Friday, November 09, 2007 9:12 PM
To: ccielab@groupstudy.com
Subject: average percent drop
Hi GS
I know this is not a really CCIE lab question, but I have been looking
around and I have not found the answer so i thought GS is a great place to
try.
i remember that i read sometime ago that there is a % of average drop in all
data links, not taking in consideration what kind of traffic you are
passing trough or QOS strategy , it seems there is some sort of threshold
we should be expecting to lose. like if we have a T1 connection, a
0.1%might be naturally drop.
by any chance any of you might have this info, or if im just thinking crazy
could someone say "yes u are crazy"
thank you guys
JC
This archive was generated by hypermail 2.1.4 : Sat Dec 01 2007 - 06:37:29 ART