|
|
| Author |
Message |
| 9 new of 83 responses total. |
janc
|
|
response 75 of 83:
| 
Jul 19 17:13 UTC 1998 |
OK, I did a little more checking. Here's two snapshots from March showing
14 dialin users on Grex:
--USER-- --LINE-- ------HOST------ ---------SINCE----------
monamoor ttyrf 204.212.46.131 Mon Mar 23 22:26:50 1998
cmcgee ttypc 204.212.46.131 Mon Mar 23 20:59:50 1998
orinoco ttyt3 204.212.46.131 Mon Mar 23 22:10:58 1998
snow ttyt4 204.212.46.131 Mon Mar 23 22:07:00 1998
illogic ttyrc 204.212.46.131 Mon Mar 23 22:26:15 1998
bye ttyt0 204.212.46.131 Mon Mar 23 22:27:12 1998
sprice ttyqc 204.212.46.131 Mon Mar 23 22:10:34 1998
tpryan ttyq6 204.212.46.131 Mon Mar 23 20:36:11 1998
sixx ttyq3 204.212.46.131 Mon Mar 23 22:27:18 1998
steve ttys9 204.212.46.131 Mon Mar 23 20:59:54 1998
n8rxs ttyp1 204.212.46.131 Mon Mar 23 20:14:35 1998
gibson ttypb 204.212.46.131 Mon Mar 23 21:18:08 1998
noreturn ttyp7 204.212.46.131 Mon Mar 23 21:26:52 1998
mjg ttyr2 204.212.46.131 Mon Mar 23 22:16:58 1998
--USER-- --LINE-- ------HOST------ --------SINCE-----------
omni ttys6 204.212.46.131 Thu Mar 26 21:41:53 1998
mcnally ttyte 204.212.46.131 Thu Mar 26 21:40:02 1998
robh ttysd 204.212.46.131 Thu Mar 26 21:23:29 1998
deigert ttyr7 204.212.46.131 Thu Mar 26 22:06:50 1998
sekari ttyq9 204.212.46.131 Thu Mar 26 22:10:34 1998
n8rxs ttytd 204.212.46.131 Thu Mar 26 21:09:21 1998
cmcgee ttyr3 204.212.46.131 Thu Mar 26 21:43:24 1998
kami ttys5 204.212.46.131 Thu Mar 26 21:23:55 1998
beamer ttys0 204.212.46.131 Thu Mar 26 21:49:17 1998
gibson ttyre 204.212.46.131 Thu Mar 26 22:13:50 1998
steve ttyrf 204.212.46.131 Thu Mar 26 22:14:08 1998
orinoco ttyr1 204.212.46.131 Thu Mar 26 21:47:13 1998
n8nxf ttypa 204.212.46.131 Thu Mar 26 21:43:37 1998
raven ttyq5 204.212.46.131 Thu Mar 26 20:43:25 1998
Every 14-user snapshot I've seen has "steve" in it, who is, I believe, the
most frequent user of the staff dial-in line. I don't know of any way to
factor the staff line out of this data. It isn't distinguishable from
anything in the wtmp log.
So basically we should consider Grex full at 13 lines, not 14 as I said before.
It's possible that at 13 lines, the staff line is in use, so there still is
one free public dial-in, but it's also possible that all lines are in use.
|
janc
|
|
response 76 of 83:
|
Jul 19 17:16 UTC 1998 |
So it looks like Grex's lines fill up for something a bit under an hour a
month, around a tenth of a percent of the time.
|
janc
|
|
response 77 of 83:
|
Jul 19 17:20 UTC 1998 |
Hmmm - I'd be tempted to suggest eliminating the staff line - except it also
doubles as a voice line for the pumpkin, something we do kind of need.
|
other
|
|
response 78 of 83:
|
Jul 19 17:34 UTC 1998 |
i would not eliminate the staff line. if it can be used for remote rebooting,
or other service to grex then kkeping it outweighs its cost.
|
scott
|
|
response 79 of 83:
|
Jul 19 18:23 UTC 1998 |
The staff line isn't hardwired into Grex, either. It's possible to connect
to Gryps, too.
|
aruba
|
|
response 80 of 83:
|
Jul 19 18:58 UTC 1998 |
Ah, I didn't even know the staff line had a modem on it - I thought it was
just for voice calls. But now that you mention it, I do remember the debate
about how the staff needs to be able to dial in when the internet connection
is hosed.
|
scg
|
|
response 81 of 83:
|
Jul 20 06:41 UTC 1998 |
If steve always uses the staff line, and nobody else does at all regularly,
could the stats just be run in a way that wouldn't count steve?
|
dpc
|
|
response 82 of 83:
|
Jul 20 20:24 UTC 1998 |
Any way your slice this *very* nice data, we could cut 3 lines and rarely
cause any busy signals.
|
rtgreen
|
|
response 83 of 83:
|
Jul 20 23:04 UTC 1998 |
that would depend on your definition of 'rarely'. Jan, I hate to be the
one to keep asking for a few more lines of code, but would it be possible
to list incidents of high utilization in detail? As the count increases
from 11 to 12, note the timestamp. As the count decreases from 12 to 11,
output a record showing begin, end times and duration. This, IMO, will
show us a better picture of the impact of an 11-line configuration: The
maximum time a user would be waiting (or attack-dialing) to get in during
the peak usage. Is that three hours/month in one peak? (unacceptable, in
my opinion to ever have to wait that long for a connection) or 6 minutes,
once a day (probably manageable)?
Realize also, that we want to serve all the users, so that we won't
simply clip the peaks off the curve. Instead, we will extend the duration
a/o frequency of peaks at 11 so that the area under the curve is constant.
This means that the times a user will get a busy with 11 lines available
is greater than the amount of time we see '11 or more' lines in use today.
|