|
|
| Author |
Message |
| 25 new of 83 responses total. |
aruba
|
|
response 50 of 83:
| 
May 10 05:12 UTC 1998 |
It seems that everyone wants to volunteer someone else to be membership
chair...
|
keesan
|
|
response 51 of 83:
|
May 10 16:03 UTC 1998 |
Maybe we don't need a 'membership chair', just various people working in
various ways to encourage or help other people to join, such as by posting
ideas in coop that other people feel they can follow up on.
|
srw
|
|
response 52 of 83:
|
May 10 16:43 UTC 1998 |
I agree with Remmers and Valerie about this.
While I support saving $40 by dropping the ICnet link, I do not support
dropping the count on the local dialups yet. I expect usage to grow
there, not shrink.
Furthermore, we may be adding limited IP service by dialup in the near
future. This would induce higher dialup usage. If we can really
anticipate continued non-use of some dialup lines for multiple months, I
would certainly acquiesce, but I am not yet convinced.
|
other
|
|
response 53 of 83:
|
May 10 16:51 UTC 1998 |
i am a frequent grexer, and i use dialup access at least 75% of the time.
I have never been informed of all lines being full.
Given Grex's stated purpose of providing access for the 'have-nots,'
eliminating dialup access would be completely counterproductive. Assuming
that potential users will already have net access by which to telnet here
would be unrealistically optimistic.
I used Grex on and off for about four years before having any other net access
i could use from home.
|
aruba
|
|
response 54 of 83:
|
May 10 19:24 UTC 1998 |
Once again, no one is proposing eliminating dial-up usage. I just think we
should have the right number of modems to fit the demand for them, and that
means we should be willing to adjust downward as well as upward.
It's not the end of the world, folks, to drop a few lines. It's not as if
we were laying people off or anything. :)
Steve, are we really going to add PPP dialup service in the next 3 months?
If so, that's the best reason I've heard so far for keeping extra lines.
|
janc
|
|
response 55 of 83:
|
May 11 19:05 UTC 1998 |
I'll put PPP-dialup service on the (imaginary) agenda for Wednesday's Grex
staff meeting. I'll report back after that meeting on how long it will take
us to bring up PPP-service and what the obstacles are. My un-informed guess
is that it can be done quickly and easily. If this is true, then maybe we
should be keeping all the dial-ins. If not, I agree that dropping a phone
line that isn't being used DOES NOT mean that Grex isn't dedicated to
maintaining good quality dial-up access in the Ann Arbor area. I absolutely
believe that dialup access should be a high priority here, but I don't think
that means we need to make charitable donations to Ameritech.
|
scg
|
|
response 56 of 83:
|
May 11 21:53 UTC 1998 |
If we drop a few lines that aren't being used, it will save money that can
then be spent on things that will be used.
|
keesan
|
|
response 57 of 83:
|
May 12 00:16 UTC 1998 |
$240 per year per phone line requires 4 paying members per line. Did you
say there were 15 lines? That would be over half the membership dues just
for phone lines, not a negligible expense. Must be a lot of extra donations
in addition to membership dues keeping grex afloat.
|
scott
|
|
response 58 of 83:
|
May 12 10:45 UTC 1998 |
There is 13 dialin lines. We have lines into the Pumpkin also for the ICnet
link and for a staff line.
|
keesan
|
|
response 59 of 83:
|
May 12 16:26 UTC 1998 |
Would it help at all to put something in the motd (or somewhere that only
dialin users would see it) explaining how much it costs to keep all the phone
lines in service, and that we would appreciate donations, especially from
dial-in users who are not members, toward the phone costs? How many nonpaying
users use the dialin lines?
|
dang
|
|
response 60 of 83:
|
May 12 18:14 UTC 1998 |
There's no way of knowing, especially now that we are on the terminal server.
Hey Mark, re way up there, I notice that you post the expenses. I even read
it, after I link it co coop. :)
|
scg
|
|
response 61 of 83:
|
May 12 20:09 UTC 1998 |
We can still tell who is coming in on dial-up lines, because we can tell who
connects to Grex from the terminal server.
I really hope we don't get into raising money by guilt tripping people. Doing
things to reduce the quality of service aren't good either. But, if we have
dial-up lines that aren't being used, it doesn't make sense to keep paying
for them while complaining about not having enough money for stuff we are
going to use.
|
mta
|
|
response 62 of 83:
|
May 13 21:13 UTC 1998 |
I don't think anyone said the need was critical -- just that we should start
lloking at options -- unused "stuff" is the obvious first step. If I could
see a possibility of a sudden influx of new dialins, I'd be against cutting
lines right now -- but we don't have any plans that I know of that are likely
toresult in a sudden influx of new dialins...
|
janc
|
|
response 63 of 83:
|
May 14 05:11 UTC 1998 |
I'm going to enter a separate item about PPP connections on dial ups. Mostly
this seems very feasible to do in the short range.
|
rtgreen
|
|
response 64 of 83:
|
May 14 05:38 UTC 1998 |
I remember hearing somewhere that 8 users per dialup line is a ratio that
the commercial services consider excellent service. A few responses ago,
I read a figure of four paying members for each of our lines. How can we
determine how many local non-member users there are as well? Since our
purpose is to serve all the users, not just the members, we need to know
how many locals we need to plan for.
I'm not adept at unix utilities, but is there someone more skilled who
can do something like this?:
1) grep the logfiles (a month or more) for the IP address(es) of the
terminal server
2) parse out the userid.
3) sort.
4) eliminate duplicate lines
5) count lines.
Do this for each individual month for the past year. Project the trend
out 3 months, 6 months. Divide the 3 month projection by 8. If less than
13, we might save money without apparently reducing service by suspending
some phone lines. If we will need those lines within three months, we'd
be better off keeping them. If we won't need those lines for 6, or even 9
months, we'll save some real dollars to spend on outreach, so we can build
our user base and bring the lines back sooner!
|
aruba
|
|
response 65 of 83:
|
May 17 23:26 UTC 1998 |
Re #57: Sindi, you're a little off on your numbers. Type "only 1;only 38"
at the next prompt to see the ones I calculated. No, the phone lines are
definitely *not* a negligible expense. If you want more information on
the level of Grex's donations, see the recent treasurer's reports. They're
all collected in ~aruba/reports, and there's also a file called 501c3 there
which summarizes our financial activity over the past 3.25 years.
|
janc
|
|
response 66 of 83:
|
Jul 18 19:58 UTC 1998 |
I wrote a program that generated stats on how many dialup lines are being
used. This works with Grex's wtmp log, so in theory it should be exactly
accurate, unlike statistics based on polling. In practice, there are possible
problems due to glitches in the log files and perhaps bugs in the program.
of the four months I collected data on, I think the first two are accurate.
There are fourteen modems, so the number of dial-in users is always between
zero and fourteen. The following tables show, for each possible number of
users, what percentage of the time exactly that number of users was logged
on to Grex.
Apr 1998
USERS HOURS PERCENT
0: 46.20 6.42%
1: 90.69 12.61%
2: 124.13 17.26%
3: 118.71 16.51%
4: 105.02 14.60%
5: 78.97 10.98%
6: 59.68 8.30%
7: 41.99 5.84%
8: 26.79 3.72%
9: 12.99 1.80%
10: 7.35 1.02%
11: 3.41 .47%
12: 1.94 .26%
13: 1.06 .14%
14: 0.00 .00%
TOTAL: 718.93 100.00%
May 1998
USERS HOURS PERCENT
0: 49.47 6.64%
1: 104.21 14.00%
2: 134.36 18.05%
3: 132.46 17.80%
4: 107.02 14.38%
5: 76.91 10.33%
6: 57.19 7.68%
7: 35.92 4.82%
8: 21.02 2.82%
9: 13.69 1.84%
10: 6.09 .81%
11: 2.32 .31%
12: 1.90 .25%
13: 1.24 .16%
14: 0.19 .02%
TOTAL: 743.99 100.00%
Note that users would only get a busy tone when dialing Grex if there are
14 users already dialed into Grex. This data suggests that that happened
for a total of 12 minutes over the two months of April and May.
There is a glitch in the log file sometime in the middle of June, including
some null entries and some other bad stuff. It causes my program to complain
bitterly, but it manages to recover and run onward. However, I believe that
it loses a log-out record somewhere in there, or more likely picks up a
bogus login, so it thinks a user has been dialed in continuously from sometime
in the middle of June to the present. That's why the period of time with
zero users is low in the June report, and zero in the July report:
June 1998
USERS HOURS PERCENT
0: 29.61 4.11%
1: 98.83 13.72%
2: 124.26 17.25%
3: 117.72 16.35%
4: 103.58 14.38%
5: 87.68 12.17%
6: 70.20 9.75%
7: 41.35 5.74%
8: 22.72 3.15%
9: 12.19 1.69%
10: 6.36 .88%
11: 3.45 .47%
12: 1.33 .18%
13: 0.62 .08%
14: 0.06 .00%
TOTAL 719.96 100.00%
July 1998 through Sat Jul 18 15:06:47 EDT 1998
USERS HOURS PERCENT
0: 0.00 .00%
1: 33.48 7.91%
2: 57.31 13.54%
3: 56.05 13.24%
4: 53.06 12.54%
5: 52.33 12.36%
6: 51.68 12.21%
7: 42.44 10.03%
8: 32.01 7.56%
9: 21.92 5.18%
10: 12.41 2.93%
11: 5.66 1.33%
12: 2.75 .64%
13: 1.41 .33%
14: 0.58 .13%
TOTAL: 423.09 100.00%
If my theory about what messed these up is rignt, then we can fix the July
report by shifting it up one row:
Corrected? July 1998 through Sat Jul 18 15:06:47 EDT 1998
USERS HOURS PERCENT
0: 33.48 7.91%
1: 57.31 13.54%
2: 56.05 13.24%
3: 53.06 12.54%
4: 52.33 12.36%
5: 51.68 12.21%
6: 42.44 10.03%
7: 32.01 7.56%
8: 21.92 5.18%
9: 12.41 2.93%
10: 5.66 1.33%
11: 2.75 .64%
12: 1.41 .33%
13: 0.58 .13%
14: 0.00 .00%
TOTAL: 423.09 100.00%
|
janc
|
|
response 67 of 83:
|
Jul 18 19:59 UTC 1998 |
My conclusion from this is that one phone line would never be missed, even
if dial-up usage increases significantly.
|
mta
|
|
response 68 of 83:
|
Jul 18 20:34 UTC 1998 |
Thanks, Jan!
It looks like we might even be able to go up to 3 lines without significant
impact if we had to...
|
scott
|
|
response 69 of 83:
|
Jul 18 21:01 UTC 1998 |
Thanks, Jan. This supports my own feel for the situation, for which
hard stats were lacking.
|
scg
|
|
response 70 of 83:
|
Jul 18 21:24 UTC 1998 |
Thanks, Jan.
|
aruba
|
|
response 71 of 83:
|
Jul 18 22:18 UTC 1998 |
Thanks, Jan. There's something wrong, though, because unless I'm behind
the times we only have 13 dial-in modems, not 14. (We have 15 total: the
dial-ins plus the staff line (7541) and the link modem (8228) which until
recently dialed out to ICNET.) Could it be that you're counting logins on
the console as if they were people dialed in?
|
scg
|
|
response 72 of 83:
|
Jul 19 02:33 UTC 1998 |
Or am I perhaps the 14th dial-in user, since I usually come in from Gryps?
|
scott
|
|
response 73 of 83:
|
Jul 19 02:38 UTC 1998 |
Could be the staff line. I do know that STeve Andre' uses it a fair amount.
|
janc
|
|
response 74 of 83:
|
Jul 19 16:38 UTC 1998 |
It is looking for telnet connections from the terminal server, so gryps and
the console wouldn't count, but the staff line would.
Hmmm...It's hard to check this code, because there is just so *much* data.
I'll try a few more things. I should be able to make it printout out a snap
shot of who is on each time the load goes above, say, 12.
|