On our PRIs we see connections the same as the customers' modems or
differing by 2k in _either_ direction.
After a series of experiments this morning, part of what's happening
when they differ is pretty obvious. The timeline is some variation on
this:
modem negotiation
modem reports INITIAL speed to user <--
login prompt
modem renegotiation to new speed
rest of login/authentication sequence
PM displays and logs NEW speed <--
...
I made about 20 calls to our PRIs this morning. About half the time
there was a 2k difference between what the modem reported and what the
PM showed/logged. Every time they differed, "show modem" showed 1
renegotiation.
So I thought I had The Explanation nailed. This is all due to modem
renegotiation during the login sequence - after the user's modem
reported a speed but before the PM did.
But then I decided to test it one more time. This time I'd use
another admin session to see if the PM displayed a speed during that
interval. It does. I dialed in, used the second admin session to
find the modem (m11) in "connect" state, and did repeated "sh m11"
until it went to state "active". The first speed it reported was 52k,
and that's what the modem reported. So far, so good.
Now try to log in. There's a long pause before the "!" of "!root"
even echos - aha, it's going to change, and I'm going to confirm The
Explanation. When I got logged in and used the other admin session to
show my speed, it was 48k - ooops, what happened to +/- 2k? Oh well,
sh*t happens. The core truth - that the speed changes - is confirmed.
So the modem reported 52k, the PM says 48k. The radius log should
also say 48k, right? Wrong. The radius log says 50k.
What else the PM showed this time is that there was 1 retrain and one
renegotiation. Here's what I think the timeline was:
modem negotiation
modem reports 52k, PM initially shows 52k
login prompt
modem retrain, new speed is 50k
rest of login/authentication sequence
PM logs radius Start record at 50k
modem renegotiation, new speed is 48k
PM displays 48k, with 1 retrain and 1 renegotiation
...
(Connect-and-run testing would have counted this as a 52k session
since that's what the modem reported. Radius log analysis of Start
records would have counted it as 50k.)
So The Explanation survives, with a bit of elaboration ... for PRIs.
It does not explain the large differences I'm seeing on my new CT1s
when no retrains or renegotiations are reported.
-- Dick St.Peters, stpeters@NetHeaven.com Gatekeeper, NetHeaven, Saratoga Springs, NY, 1-800-910-6671 (voice) Saratoga/Albany/Amsterdam/BlueMountain/Cobleskill/Greenwich/ GlensFalls/LakePlacid/NorthCreek/Plattsburgh/... First Internet service based in the 518 area code - To unsubscribe, email 'majordomo@livingston.com' with 'unsubscribe portmaster-users' in the body of the message. Searchable list archive: <URL:http://www.livingston.com/Tech/archive/>