Maintenance modems and power failures
Sean Donelan
SEAN at SDG.DRA.COM
Thu Feb 11 05:42:01 UTC 1999
tex at shrubbery.NET (Austin Schutz) writes:
> Get two modems.
>
>POTS ---- Modem 0 ---- Modem 1
> | |
> UPS AC Utility AC
>
> This only really makes sense if you have your terminal server
>w/ redundant power, one utility AC and the other UPS AC.
I've had troubles with two auto-answer modems daisy chained off the
same pots line, but a good solution for some cases.
A few people asked how I currently solved the problem. I'm using
three different methods depending on the importance of the facility.
1) Nothing, and praying the modem answers
Obvious problems
- You name them
2) A Digital POWERswitch, which is a cool box containing essentially
an automatic transfer switch with two AC power cords for a piece of
equipment with one AC power cord (e.g. modem, switch, hub, etc)
Obvious problems
- Trying to find a Compaq salesperson who will let you order one
- A big bundle of wire stuffed into the corner of the rack
- People get confused, and unplug the equipment from the POWERswitch,
and plug it directly into the wall outlet. I don't know why, but
for some reason people look at it and think "That can't be right,"
and zap, there goes my modem.
3) "Network monitoring systems," which is a self-built Alpha unix box,
with internal 14.4 modem (I've found them more reliable, under more
conditions than 28.8 or 33.6. I don't need speed, I need a connection.),
8 serial ports to connect to the equipment consoles, and DC power supplies
connected to the dual A&B battery supply. Unlike AC power cords which
people unplug at will, DC power scares enough people so they don't
touch it.
Obvious problems
- Not labeled by UL, FCC, or NEBS; some building engineers object
to anything without approved labels
- DC powered, Ok for telco, but a problem other places
- Vendor field maintenance not available, self-built
- More complicated systems fail more often
None are perfect, and each can fail in different interesting ways.
Now, back to your regular network operations discussions.
--
Sean Donelan, Data Research Associates, Inc, St. Louis, MO
Affiliation given for identification not representation
More information about the NANOG
mailing list