>On this topic of disconnects and all, there's a lot of info in the
>syslog, and events last week made it clear we needed to be looking at
>it, so I have a script to do that now. It will do a bunch of things,
>even hour-by-hour analysis of terminations:
> ftp://ftp.netheaven.com/pub/isp/discon
>
>It's a shell script, so it's only for unix folks.
>
>Anyway, here's why we needed it. Something odd happened over the last
>week. Have a look: here are today's termination stats so far, on our
>PRIs from a CLEC and from Bell Atlantic:
> CLEC:
> 5.49% call-circuits-closed
> 8.39% abnormal terminations, all kinds
> BA:
> 4.10% call-circuits-closed
> 7.46% abnormal terminations, all kinds
>
>Last Thursday's terminations on the same PRIs:
> CLEC:
> 16.05% call-circuits-closed
> 21.69% abnormal terminations, all kinds
> BA:
> 10.65% call-circuits-closed
> 17.19% abnormal terminations, all kinds
How many calls are those stats based on ? Here are mine for a week on some PM3s
normal 85.76% 42055/49039
User Request - Call Circuit Closed 7.06% 3461/49039
Lost Carrier 6.41% 3145/49039
User Error - PPP NCP Active to Request 0.50% 244/49039
Port Error - Exceeded LAPM retransmission limit 0.12% 61/49039
User Request - PPP Term Ack 0.05% 23/49039
NAS Request - PPP Maximum Retransmissions 0.04% 18/49039
Service Unavailable - PPP No Protocol 0.02% 8/49039
Service Unavailable - Host 0.01% 7/49039
Service Unavailable - Master out of Sync 0.01% 7/49039
Service Unavailable - Failed to detect V.42 remote 0.01% 6/49039
No Event Identified 0.00% 2/49039
User Error - LAPM negotiation timeout 0.00% 2/49039
The Lost Carriers are kinda high, but its partially based on the fact
that we have a LOT of Windows 3.1 SLIP users as we are one of the
only people in the area to offer support for those folk.
>What's really curious is that it cleared up on the CLEC PRIs first;
>the BA PRIs didn't improve until today. At first I wondered if there
>was something the CLEC hadn't told me, something they'd done to clear
>up their lines. However, I don't see how they could have cleared up a
>problem with BA calls to BA PRIs in another city ... the CLEC's off
>the hook on this one.
Software upgrades perhaps ?
>
>I ran an hour-by-hour analysis of terminations all the way back
>through August and got a surprise. Last week's problem wasn't new, it
>was just suddenly a whole lot worse. The disconnects have been off
>and on since August, and they have averaged much higher than I would
>have guessed based on client complaint level. I suspect both we and
>our clients have been tending to see disconnects as v.90/flex teething
>problems.
Depending on your area, it could be the way that calls are routed at peak times..
Also, usage patterns can really be odd... Our power users tend to be on
at off hours, and they are the ones generally with good equipment. Our non
tech users-- the ones with crappy modems -- tend to cluster around
peak hours, so that might account for some of the differences for periods.
In short, there are so many potenial variables at work, its practically
impossible to pin it down...
>
>However, it's clear that we've been at the mercy of gremlins somewhere
>deep in the BA network ... causing lots disconnects that weren't due
>to our lines or our clients' lines or their modems or our modems or
>v.90 or our ComOS or our CLEC's whatevers ...
v.90 disconnects show up as Call Circut Closed quite often...
Mike Tancsa (mdtancsa@sentex.net)
Sentex Communications Corp,
Waterloo, Ontario, Canada
-
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/>