|
|
| Author |
Message |
| 25 new of 270 responses total. |
jep
|
|
response 150 of 270:
| 
Jul 10 01:12 UTC 1995 |
Grex is welcome to teach people how to use .forward files, and have
them run POP on M-Net. They will have to log in to their account every 30
days or more to avoid being reaped, a POP session doesn't count.
We currently have between 25-50 users who use POP, among 6000 +/-1000
accounts on M-Net.
|
robh
|
|
response 151 of 270:
|
Jul 10 02:12 UTC 1995 |
Question - Do you have to be a member or patron to use the M-Net
POP server? I had a help-seeker tell me he couldn't get Eudora
running with his M-Net account.
|
mdw
|
|
response 152 of 270:
|
Jul 11 06:10 UTC 1995 |
I'm beginning to suspect we're getting a fundementally different sort of
internet user than m-net. I think our name of "cyberspace.org" is
attracting a significant number of people who are kind of internet
savvy, but don't really have any idea what grex is. This group of
people seems to have little understanding (or interest) in conferencing,
but seems to think of grex as a convenient ftp site, web site, neat
e-mail address, and so forth. I suspect many of these people think that
grex is somehow supported by the government, or by their ISP - and I
would guess that these people don't realize how pathetic grex's internet
connection is, because they're sitting at the other end of a slow 28.8k
link so any lack of response on grex's part is much less obvious.
It seems like the number of people who are asking about POP has been
sharply rising of late. My guess is we could easily see 60-100 users
using POP, in not much time at all. If that was just personal mail,
that might still not be that big a deal, but a considerable fraction of
our mail seems to consist of mailing lists.
|
jep
|
|
response 153 of 270:
|
Jul 15 08:15 UTC 1995 |
No, you don't need to be a patron or member to use POP on M-Net. If
people are having problems, please have them send mail to
"staff@arbornet.org", we will try to help them. Thanks!
|
sidhe
|
|
response 154 of 270:
|
Jul 23 19:51 UTC 1995 |
I don't think the apprehensiveness regarding initiating POP is not
necessary. There is, indeed, such a thing as trial-and-error. If you try
the idea out, and there's major problems, then rip it out, and throw it
away, at least being richer for having tried it.
|
lilmo
|
|
response 155 of 270:
|
Aug 22 06:02 UTC 1995 |
Or, if it doesn't work out at first, make it a members-only option. If it is
still a problem (unlikely, since we don't, I think have enough members now
to MAKE it a problem, and if we got enough, that would be a sufficiently
large boost to our bottom line that we could make it NOT a problem) THEN
rip it out.
|
ajax
|
|
response 156 of 270:
|
Sep 25 21:26 UTC 1995 |
I have a suggestion I've been meaning to make for a few months.
How about lowering the maximum number of users who can be on Grex
at once? I don't know the current limits, but it seems just from
experience like if there are somewhere around 45 folks on Grex, it
plugs along moderately tolerably, but when the number is closer to
50, it really bogs down.
I know the Sun 4 is coming soon, and that will change the balance,
but if this is a quick and easy change, I think it's worth considering
just for the short term, until the Sun 3 is relieved of duty.
|
steve
|
|
response 157 of 270:
|
Sep 25 22:10 UTC 1995 |
It's almost a no brainer to do this. Just zap some of the pty's
that we're down.
The problem is, what do we ramp it down to? Grex usually is
sort of bearable until it gets 30 people on. After that things
start to go downhill fairly quickly. But, if we aren't processing
mail, and most people are just in party or bbs, the system isn't
too slow with 40. When a sendmailstorm hits us, even 20 people
during the storm is kinda bad.
So how do we come to a limit? If we're consistently hitting
upwards of 45+ people, does that mean that enough peple still like
Grex (even though as fast as a glaicer), such that we'd be doing
the wrong thing in limiting the users?
|
ajax
|
|
response 158 of 270:
|
Sep 26 04:51 UTC 1995 |
Clearly there are people willing to put up with the slowness,
as you point out. However, it filters some people and activities
on Grex, in ways that in my opinion aren't desirable.
One example is that once party is started, I gather it responds
reasonably well when the system is slow, while Picospan doesn't.
(Earlier today it was taking 30 seconds to display each item).
So if the max users is set to 40, you might get five people in
bbs and 20 in party, but if it's set to 50, you might get two in
bbs and 30 in party, tilting the balance away from conference
participation. If you view all uses of Grex as equally valuable,
then having the most people serviced at levels they tolerate is
best; but frankly, I don't :-).
Another reason I think the balance isn't good is that I think
the people who are more likely to contribute to Grex (in money or
time) are often more likely to be more interested in conferences
than in party, and when things are really slow, these people will
lose interest in Grex.
As to your question, where to set the value, I realize it's an
arbitrary decision, but I'd say lower it by between five to eight
ptys. It's just my preference, but I thought I'd throw it out there
and see what other people think.
|
rcurl
|
|
response 159 of 270:
|
Sep 26 06:15 UTC 1995 |
Could the limit be tied to the load?
|
lilmo
|
|
response 160 of 270:
|
Sep 26 12:00 UTC 1995 |
I'm guessing no, but it would be neat if we could. OTOH, what happens if the
limit is on, and suddenly the load goes up (sendmailstorm), and the limit thus
goes down: would ppl get kicked off, or just no more would be able to get
on until the # on gets below the limit?
|
steve
|
|
response 161 of 270:
|
Sep 26 14:58 UTC 1995 |
Five to eight pty's? I think we'd have a revolt on our
hands. We have 48 of them, so we're talking about an 80%
cutback. Ouch.
Could we limit the number of lines? Well, yes we could. My
first thought would be to create an daemon that sits there and
creates ptys depending on the load. But it would also have
to decrease them, and once the pty's were increased the load
would go up, and start the cycle all over again. I think
that if we're really serious about this, making a flat
number of them would be best. Like 32.
|
selena
|
|
response 162 of 270:
|
Sep 26 17:27 UTC 1995 |
Are you NUTS?
<sorry, just a little technical consideration>
This place is hard enough to log into already, with all the crashes!
And, after waiting gods-know-how-long to get through to michnet, you
mean I might have to wait another hour or better, just so that
things are a little *faster* when I _do_ get here? I suppose
this means that once I get on, and I look at the who list, and
don't see half the people I'm here to talk to, I'm out of luck?
I mean, essentially, I'd have to wonder if *my* tty
would have been the one they would have gotten, if their attempt
was just a little better timed..
And what do i do, ask people to get off, so my friends can
get on? Or, encourage others to leave? Look, I thought this was supposed
to be a community.. limiting the amount of users for the sake of
performance only works for those who care more about the machine's
functions than the people using it. I come here for the people, as
do many others.
Hell, while you're at it, just dump the link! You'll find it
runs REALLY fast, then..
It's like saying "this party wouldn't be so crowded if we
didn't let as many people in". Of course it wouldn't be. And who'd
want to attend?
|
ajax
|
|
response 163 of 270:
|
Sep 26 19:16 UTC 1995 |
STeve, maybe you misunderstood? I meant cut the max by 5 or 8 ptys,
not set it *to* 5 or 8 ptys. I.e., if the current max number of users
is 52 or so (that's about the highest I've seen it), maybe try it at 45
or 47. It seems like adding the last 10% of users means that everything
runs at about one fourth the speed.
Selena, if you want an easier time getting into Grex and having lots
of people on, I think an always-just-tolerable level is a short-sighted
alternative to a mostly-satisfactory level. If people can dial up Grex
and expect pretty reasonable service in the evenings, I think we'd reverse
Grex's trend of declining memberships. If the Sun 4 works as planned,
we'll want to increase the ptys again, but again we'll be faced with the
same question: should we again increase it to the point where the load
average is usually in the double-digits during prime time?
|
steve
|
|
response 164 of 270:
|
Sep 26 21:41 UTC 1995 |
AH---
Thanks Rob. Maybe I need to have my head realigned?
Deleting 5 or 8 won't really solve the problem, because
from what I've seen, things start to get bad once we've
hit 30 people or so. Thats why I suggested 32 pty's;
it would make for a system with 43 available normal
ports on it.
|
rcurl
|
|
response 165 of 270:
|
Sep 26 22:51 UTC 1995 |
The problem is to spread users out more evenly. Selena, try around
5 a.m. (Ann Arbor time).
|
steve
|
|
response 166 of 270:
|
Sep 26 23:12 UTC 1995 |
Well, that would work. It's impossible to do, but it would
work.
|
rcurl
|
|
response 167 of 270:
|
Sep 27 06:12 UTC 1995 |
It just takes some inspirational hype:
DON'T MISS OUT ON THE FUN
HAPPY TIME ON GREX
## EVERY MORNING ##
4- 6 AM !!
NO LAG !!
|
srw
|
|
response 168 of 270:
|
Sep 27 07:48 UTC 1995 |
STeve and Rob, I think everyone's definition of slow is different.
Rob thinks the last 8 ptys make a big difference. I tend to agree.
Even though STeve is right that at 30 it starts to get slow.
Knocking off 16 as you propose, STeve, is too big a change to make, IMO.
We currently max out the ptys on Grex only for a short period every day.
If we dropped the ptys by even 5, that amount of time we spend
maxxed out would be substantially greater. I wouldn't want to drop
more than that, at least as an initial experiement.
|
robh
|
|
response 169 of 270:
|
Sep 27 10:11 UTC 1995 |
I remember all too well when we were down at 24 pty's, was
that a year, year and a half ago? (I said I remember it, I
didn't say I remembered the date. >8) It was just about
impossible for anyone to get on, and we got more complaints
about lack of access than we ever did about our slowness.
(I know I got those complaints, on those rare occasions when
I could connect...) I would rather keep the pty's open,
and focus on getting a faster system.
|
sidhe
|
|
response 170 of 270:
|
Sep 27 14:16 UTC 1995 |
I'm with robh and selena. Making access more difficult will not
endear you to new membership, and how could it, when access is everything,
on an interenet-linked system? You already have a limit on the number of
terminals that can access grex at once, by sheer load limit. Many people
come here for other people, it is true, and if you diminish that one
resoure that sets grex apart from nether net or m-net, you are only
hurting yourselves.
|
ajax
|
|
response 171 of 270:
|
Sep 27 14:27 UTC 1995 |
Re 167: <g>! You sure you aren't in advertising, Rane?
Rob, if you keep a "have max ptys open, and focus on a faster system"
philosophy once we get a faster system, it won't be any faster! :-)
Now is as good a time as any to maintain a good balance, and it will
surely need adjustment with the Sun 4. I've come to agree reduction
by 8 right off would be high, but I think reducing even two or three
would be a useful step in the right direction.
|
ajax
|
|
response 172 of 270:
|
Sep 27 14:40 UTC 1995 |
Re: 170...in answer to how it could draw new members, my impression
(and correct me if wrong!) is that a much higher ratio of direct-dialers
become members than telnetters. Reducing ptys would make Grex more
attractive to them. And when people do get on and it's slow, I think it
attracts more people who will put up with it because it's free, but
wouldn't want to pay for it.
I share your concerns about getting more people on, I just have a
different approach getting there. With more members, we could afford
to improve Grex, and ultimately allow many more ptys.
|
scg
|
|
response 173 of 270:
|
Sep 27 14:57 UTC 1995 |
Maybe I'm just being impatient, but it's gotten to the point where I often
don't have the patience to use Grex, due to the extreme slowness. It seems
like knocking off just a few ptys would make a somewhat significant difference
in speed, while still allowing lots of people to telnet in. Furthermore, if
Grex were faster, people would be able to get through the conferences faster,
and not have to spend as much time here. I don't share "selena's" concerns
about not being able to find people on Grex, because not only would we still
be allowing a lot of people on, but for something like Picospan, Grex's main
reason for existance, it really doesn't matter whether people are on at the
same time or different times. Party, which is also a somewhat important part
of Grex, works much better when it's small, since when a lot of people get
into party the conversations become pretty hard to follow.
|
steve
|
|
response 174 of 270:
|
Sep 27 15:46 UTC 1995 |
Steve, it won't work that way, just knocking off a few of them.
This all gets into what people are doing obviously, how much memory
there is, and what people are doing. The reason Grex is able to do
as well as it does currently is because of something called hardware
context switches, which hold information on processes. There are
eight of these on the 3/260 CPU (16 on the 4/260) and this helps
out a whole lot in terma of the system being able to keep up and
switch from one process to another.
Around the 30 person mark we seem to lose it, in terms of
speed. Below that, and we're usually fairly OK; above that
and we rarely are.
This is why I say that taking just a few of them off isn't
going to do good. The number of users on Grex does not present
a linear load on the system.
|