Grex Coop10 Conference

Item 33: Dispose of the Sun 670?

Entered by dpc on Thu Sep 11 20:55:52 1997:

It has been many, many months since the Sun 670 arrived, and it still
isn't running.  This is no fault of the staff--the darn thing is,
as one person pointed out, "a beast."
        Would it be better to admit defeat and sell the parts for what
we can get for them?
137 responses total.

#1 of 137 by dang on Thu Sep 11 21:05:46 1997:

No.  It looks like all the major technical problems are solved, and now the
only things remaining are finding time to install software, and operating
systems and whatnot.  There'd be little point in giving up now.


#2 of 137 by scott on Fri Sep 12 01:51:46 1997:

 Heck, why don't we just give up and liquidate the entire Grex 
operation?


#3 of 137 by bmoran on Fri Sep 12 13:37:35 1997:

I'll bid $5.00(US).


#4 of 137 by valerie on Fri Sep 12 21:40:39 1997:

Dave, there actually is some progress on the 670, it's just slow.
Until recently it wouldn't recognize its disks.  Mike McNally fixed that.
Now it hangs when it does the file system check as it boots.  That's the
next hurdle in getting it configured and running.


#5 of 137 by richard on Sat Sep 13 20:25:17 1997:

Grex has over $4,000 in the bank.  It could afford to swap out the 
Sun670 on an even faster model, one that would work better with Grex.  
Must be a place somewhere that would take a 670 as a partial trade-in?

If it is such a problem to ever use this 670 right, maybe it should be 
replaced?


#6 of 137 by aruba on Sat Sep 13 21:32:12 1997:

Actually, we have $3,499.84 in the bank, after paying for the routers.


#7 of 137 by senna on Mon Sep 15 06:15:49 1997:

That's a fairly hefty investment that runs a bit of an unpleasant risk.


#8 of 137 by jared on Mon Sep 15 23:09:51 1997:

$3k gets you a computer a lot faster than the 670.
but that's not the point here.


#9 of 137 by mdw on Tue Sep 16 01:06:15 1997:

Actually, that's just it.  We also want something that's reliable,
compatible with what we have now, does the things we need, and cheap to
run and expand.  The 670 is an MP machine, which means it has 2
processors.  That means even though it's slower than a uniprocessor
machine, it can do more.  We care a *lot* more about throughput than we
do about speed.  Most of the binaries we have on the sun-4 will run
unchanged on the 670, and we should be able to just unplug the current
/a partition, possibly change one dip settting, and plug it on on the
670.


#10 of 137 by dang on Wed Sep 17 22:28:48 1997:

And we can even add two more processors, although that probably would improve
things much unless we also went to solaris.


#11 of 137 by senna on Thu Sep 18 02:08:42 1997:

You mean, would not?


#12 of 137 by srw on Fri Sep 19 01:40:50 1997:

i am sure he meant "would not". Actually there probably would be some 
benefit to running a second processor even under SunOS, just not as 
much.


#13 of 137 by mdw on Fri Sep 19 15:29:37 1997:

SunOS doesn't have a threaded kernel; so doesn't support symmetric
multi-processing.  SunOS does have some support for multi-processing,
with only one processor in the kernel at a time.  On "average", Unix
sytems like grex spend about 50% in the kernel.  Therefore, 2 processors
works well, but more than 2 processors is a waste.


#14 of 137 by dpc on Sat Sep 20 20:29:06 1997:

I'm glad to see some progress is being made.  Obviously getting the
ISDN line working is a higher priority, but it looks like that should
be done in a week or so.  (hope, pray!)  
        Can anyone tell when the 670 should be ready for use?  


#15 of 137 by valerie on Sun Sep 21 13:29:46 1997:

Alas, nope.

Relax?


#16 of 137 by steve on Sun Sep 21 17:34:58 1997:

   Dave, we've gone over this before.

   We aren't going to commit to a time table because we all have other
things to do in our lives, like work, sleep, and maybe a few other things
that aren't Grex related.

   Yes--progress has been slow.  Would that it were not.  Progress was
slow on getting to the Sun-4, too, but when we got there, we were on a
stable platform that was secure.  So it will be with the 670.  I would
*FAR* rather take a year to properly come up on a new platform than 
rush into something only to have it be a disaster.

   One of M-Net's operating system upgrades immediately comes to mind.
Within days (hours?) of the new system in place, a dear little vandal
found some hole in the system (one of the rdist problems, I believe)
and proceeded to do a "rm -rf /" on the system, wiping out every file
on the system.  I remember that, as I was on the system at the time,
and saw various pieces of the system evaporate.

   Grex staff has always, always, always been over cautious when working
with things, precisely becuase if we screw up something, we've not only
wasted the original time in making the screwup, but we get to spend the
additional (probably larger) amount of time doing it *again*, to fix the
the first problem!

   Lastly, I will point out that just selling the 670 without having
a clear idea of where to go to next would be insane.  And, given that
any campaign to dump the 670 and move to an Intel box of some sort
would mean 1) Switching operating systems, 2) recompliling all of
Grex's special software, 2) Testing an entirely new operating system
for security problems.  It took about 14 months for us to switch from
the Sun-3 to Sun-4 platforms.  Moving completely from this architecture
to an Intel based monstrosity would take even longer.

   Has the progression of getting to the 670 moved at the speed of a
wounded snail?  Hell yes.

   Has Grex managed to upgrade from platform to platfom?  Again, hell
yes--there was 1) the move from the Sun-2 to Sun-3, 2) the move from
the Sun-3 to the Sun-4, and 3) currently the move from the Sun-4 to
the Sun-4/670.  In several ways this is just as hard, since its the
top-of-the-line older Sun stuff, which most of the staff hasn't had
as much flight time with, since this was the *really* expensive stuff
of its day, moreso than the random Sun-{2,3,4} stuff we've used in
the past.

   We will get there.


#17 of 137 by senna on Mon Sep 22 05:10:57 1997:

If STeve and I are talking about the same thing, then I remember the software
upgrade he's talking about.  At any rate, I remember 15 months ago when m-net
did a major upgrade.  A catastrophic crash occured, rendering m-net unviable
for close to a month, with few exceptions.  It was rarely up, and a large
number of mnetters left or came to grex.  M-net's usership took months to
recover, and I don't think it ever really got back under its feet financially
(though it was certainly reeling before then, too).  I woulnd't want that to
happen to grex.


#18 of 137 by n8nxf on Mon Sep 22 13:07:58 1997:

The world is *FULL* of complainers and most of those complainers do not
have the wherewithall to do anything about what they compain about, or
much else.  At least that's been my experiance when I ask a compliner to
help out.  Sometimes you get lucky thoug :-)



#19 of 137 by dpc on Mon Sep 22 18:19:07 1997:

As long as the snail is still moving along, can anyone predict when
it will reach its goal?  Six months from now?  A year from now?
        Obviously security is a major goal.  However, getting stuff
operational which was paid for by a fundraising drive quite some
time ago should also be a major goal.


#20 of 137 by mta on Mon Sep 22 18:45:08 1997:

It is a major goal, Dave.  Folks are working on it as fast as they can
whenever they can.  Time is one of the path elements that can't be changed
much in an all volunteer, high -skills project.  I think most everyone else
here understands that.

Do you want your money back?  Is that the point you're working up to?


#21 of 137 by steve on Mon Sep 22 21:43:44 1997:

   I'll stick my neck out Dave, and say six months or less.  Plan
on Six months, like when the ice of winter '97 is thawing.


#22 of 137 by bmoran on Tue Sep 23 12:56:16 1997:

"fundraising drive quite some time ago". That time ago was the July board
meeting, two months ago. A mere blink of the eye, in grex time. 


#23 of 137 by aruba on Tue Sep 23 20:10:37 1997:

(Dave is referring to the fundraising drive for the new computer, which ran
from Januray through March of this year.  You're thinking of the ISDN drive,
Bill.)


#24 of 137 by tao on Tue Sep 23 22:02:02 1997:

re 19: When I donated funds to the ISDN drive, it was with no
condition on how soon staff had to implement installation.

When it comes to shared computing environments such as Grex,
only a fool would rush an upgrade job.  The folks we have
on staff here bhave been around computers long enough to
know what they're doing.  I've been around computers long
enough to know that the staff aren't blowing smoke when 
they tell us it'll be a few months.

Security and stability win out over fast implementation
any day.


#25 of 137 by valerie on Wed Sep 24 01:02:32 1997:

("Natalie Drest"?)


#26 of 137 by bruin on Wed Sep 24 12:54:18 1997:

I believe I remember "Natalie Drest" as one of the people in the closing
credits of "A Prairie Home Companion."


#27 of 137 by tao on Wed Sep 24 15:25:39 1997:

(Um, yeah. Along with Norman Conquest, Immanuel Transmission, etc)


#28 of 137 by dpc on Wed Sep 24 15:37:31 1997:

When the ice thaws next spring sounds like a good goal to me.
No, of course I don't want my money back!   8-)


#29 of 137 by arthurp on Thu Sep 25 00:19:35 1997:

Thank you staff for all the work and frustration you have endured on the 670.
I've been following the progress and sympathyzing.  If I knew anything about
SunOS I'd offer to help, but I'd be a hinderance on this one.


#30 of 137 by mta on Thu Sep 25 15:15:25 1997:

The only "help" I can offer is moral support, pop, and cookies.
Let me know about the next group work session and I'll be there with 
baked goods!


#31 of 137 by steve on Thu Sep 25 21:33:31 1997:

   Now, thats a thought...


#32 of 137 by other on Wed Nov 19 07:53:32 1997:

out of curiosity, can we get an update on the upgrade progrees?
thanks.


#33 of 137 by janc on Wed Nov 19 17:17:42 1997:

Next step is to install SunOS on the machine.  We are currently running an
old version of SunOS on Grach, and we can't copy over the version from Grex
because the software for multiprocessor architectures is slightly different.
So we have to do a fresh installation from the CD.  We have two CDs.  One is
the SunOS 4.1.3_U1 version that we are running here.  Greg has it, but can't
find it, and doesn't think it includes the multiprocessor stuff anyway.  The
other is a SunOS 4.1.4 CD that Jared donated.  STeve has it, it's in his car,
but it hasn't made it to the pumpkin in the last few weeks.  When and if a
installation disk appears, someone will have to actually do the install and
copy over Grex's software.  Odds are, this is me.  I expect to be able to
do most of it in a solid days work, though there are a few sticky issues with
libc that may take a bit longer to figure out.  After that, it will mostly
be a question of doing a good security sweep of the system.


#34 of 137 by janc on Thu Nov 20 20:12:02 1997:

STeve got the CD to me after the board meeting last night.  I've done the base
install, and will be doing the Sun patches, then the Grex mods soon.


#35 of 137 by tsty on Fri Nov 21 04:13:45 1997:

yippiyyippiyyippiy-o-day!


#36 of 137 by valerie on Fri Nov 21 15:13:07 1997:

Yay Jan!  Cool beans!


#37 of 137 by dpc on Sat Nov 22 18:02:53 1997:

Great!!


#38 of 137 by janc on Sat Nov 22 21:59:45 1997:

I think I've completed the first pass of software installation on the 4/670.
Nearly all of Grex's custom software has been copied over and installed.
We still need to do our kernal modifications and a whole lot of testing.


#39 of 137 by srw on Sun Nov 23 22:31:36 1997:

I spent Sunday afternoon working on the kernel configuration for the 
4.1.4 kernel to meet Grex's specific needs, and working on the 
installation of the kernel blocks. This is complete now, and the machine 
has been rebooted to run on the new kernel but there are still a number 
of other things to take care of before we can begin the final testing 
and security checks. We have to figure out where to put the mail spool 
for one thing. 

This was definitely a good weekend for progress on the 670. I'm not at 
all sure how long the remaining work will take.


#40 of 137 by janc on Sun Nov 23 22:35:22 1997:

Steve Weiss has rebuilt the 670's kernel, so it now has all the latest Grex
mods.

What's basically left to do now is a variety of minor software work (like the
latest version of Backtalk isn't there yet), a lot of testing to make sure
that all the programs work OK, and some double-checking to ensure that all
our security is correctly installed.

Then we have to generate our change-over plan.  A lot of things on Grex are
continously changing, so they cannot be moved to the new system until we
actually make the change.  To make sure we get it all, we need to make a plan.


#41 of 137 by remmers on Mon Nov 24 02:25:40 1997:

I suggest that the move not take place during the board election
(i.e. December 1-15).


#42 of 137 by tsty on Tue Nov 25 00:14:27 1997:

fine suggetion, agreed.


#43 of 137 by lilmo on Tue Nov 25 23:08:16 1997:

I presume Grex will have to be shut down during the actual changeover?


#44 of 137 by valerie on Tue Nov 25 23:14:26 1997:

Yes, probably for between half a day and a full day.


#45 of 137 by janc on Wed Nov 26 00:15:56 1997:

I'm not exactly sure why doing the change-over during the election period
would be a problem.  As long as we move over the vote directory, it should
not interfere with the election.


#46 of 137 by mary on Wed Nov 26 00:55:21 1997:

If all goes well.  Is there any chance at all that changing 
over could result in Grex being unavailable for a couple days
or longer?  It would be nice if the vote could come off, as
planned, without the need for extensions and added confusion
as to deadlines.


#47 of 137 by valerie on Wed Nov 26 15:30:39 1997:

Ya, I'd rather not do the changeover during the voting period.


#48 of 137 by janc on Wed Nov 26 16:22:27 1997:

I think the chances of that are small.  We aren't erasing the old system
when we bring up the new system.  If there are any snags in bringing up the
new system, we can always fall back to the old system.  This assumes that
we don't screw up badly enough to leave both systems un-workable.  This is
possible, of course, but pretty unlikely.


#49 of 137 by remmers on Wed Nov 26 17:47:49 1997:

If the new system is up for a little while, some people cast
votes, and then we have to go back to the old system because of
some unforseen problem, the votes that were cast on the new
system could get lost. Unlikely perhaps, but possible.


#50 of 137 by janc on Wed Nov 26 20:03:23 1997:

Well, we are basically ready to switch except that we need to do a security
sweep and probably a bit more testing.  There is no particular reason that
this couldn't be done before December, but there are only a few people really
qualified to do a good security sweep and it's a matter of one of those people
finding time.  If we don't make it before, I guess we can wait till after.


#51 of 137 by srw on Wed Nov 26 20:52:00 1997:

I'd hate to have the new computer sit around unused for a month, once it is
ready to be used. 


#52 of 137 by dpc on Wed Nov 26 22:37:48 1997:

I'd rather see the 4/670 wait until *after* the voting period.  And remember
I've been one of the people jumping up and down about delays!
        I (and others) have suffered through several M-Net upgrades which
turned out beneficial in the long run, but caused maddening delays/
service outages/whatnot in the meantime.  We may *think* a System
change of the magnitude we contemplate will go smoothly.  But I've
been down that trail of forlorn hopes too many times.


#53 of 137 by other on Thu Nov 27 05:31:13 1997:

I'm extremely confident that the changeover would be accomplished without
significant delay or outage, but I think I agree that it makes sense to hold
off on the switch until after the election, just for the sake of
organizational expediency.  Besides, I'm a candidate and I'm terribly curious
to know the outcome <grin>.

Seriously, though, the staff has done this type of major upgrade on Grex a
couple of times in the last few years, to the Sun 3, then to the current
machine, and I'm sure that between the lot of them, they've covered almost
every possible contingency which would result in any major problem.  You folks
are incredible!  But the extra delay won't kill anybody, and it just makes
me all warm and fuzzy to think of the election going of in a timely fashion
followed by starting the new year on the new machine.


#54 of 137 by aruba on Thu Nov 27 07:04:10 1997:

Yeah, I agree with John, et. al., that we ought not to interrupt the vote.
I hope this doesn't break your momentum, Jan!  But we've waited 10 1/2 months
already; we can afford to wait another two weeks.


#55 of 137 by davel on Thu Nov 27 14:51:52 1997:

Even if there are no problems, we're likely to have a good part of a day of
downtime in the changeover, right?  Not good in the middle of a vote.


#56 of 137 by srw on Sun Nov 30 07:29:54 1997:

The total downtime expected is probably around 8 hours, but of course it 
is not easy to predict. Presumably it would happen on a weekend, when 
the load is lightest.

Well, I did my part last weekend to ready the 670 for use, though I have 
not spent the most time of any staffer. If there is going to be no more 
progress on it until January -- I'll be honest here, I wouldn't have 
bothered if I had known that it was going to sit around for 5 weeks 
after I got done. I felt pretty strongly that the Grex supporters had 
pumped a lot of money into this machine and deserved to get to use it 
ASAP. Further, I can see clear evidence that we are hurting badly due to 
our current machine's inadequacies. When we do switch over, the same 
risks will be there affecting other activities if not the vote. 
Something is always going on on Grex. In other words, I disagree with 
you all about this, but I'll just go away and work on something else.


#57 of 137 by aruba on Sun Nov 30 12:52:27 1997:

(The election ends December 15th, right?  So we don't need to wait until
January.)


#58 of 137 by other on Sun Nov 30 15:30:08 1997:

Steve, we're not really talking about a very long delay.  I'm surprised that
you're taking this view.  i think the only reason those of us who advocate
the delay do so is because of the fundamental operational importance to Grex
of timely board elections, especially at a time when the organization is
likely to come under review for 501(c)3 consideration.  I most definitely
appreciate all the hard work you and the other members of the staff have put
into preparing Grex for the move, and it is not at all my desire to delay it.
I just think it is a reasonable and sensible precaution.


#59 of 137 by srw on Sun Nov 30 21:00:07 1997:

I was thinking it would take us through December, So it is not as much 
of a delay as I had realized. It is up to someone else to do a security 
sweep on the OS, and then there is still a matter of finding a disk to 
hold the mail spool. Presumably that could be done in a few days, but 
it's my worry that no one is going to work on this for a while. It may 
be a simple precaution you are asking for, but it demotivates working on 
the system.

I jumped on the project, pushing a few things out of the way, with no 
idea that it would then get put on hold. That was the cause of my 
resentment. I was under the impression we really wanted this project 
ASAP. You all blindsided me by asking for a delay. That's why I reacted 
as I did.  

It now looks like a bit of an overreaction. I honestly hope that the 
security and other issues are dealt with in the next week or so, so that 
we can go live with the 4/670 on December 16.


#60 of 137 by lilmo on Mon Dec 1 04:20:39 1997:

Thank you for being big enuf to admit you over-reacted a bit.  I completely
understand how you felt:  I've put in massive effort to get something done
by a deadline that was then pushed back.  It kinda stings at first, but as
long as you don't nurse it, the resentment dies a lonely, painless death. :)


#61 of 137 by aruba on Mon Dec 1 08:33:59 1997:

I have been there too, and I'm sorry so see it inflicted on Steve & Jan and
the rest of the staff.  (Frankly, I didn't realize you all were that close to
being ready.)  I hope we can switch over as soon after the election as
possible.


#62 of 137 by n8nxf on Wed Dec 3 17:33:46 1997:

The work had to be done sooner or later.  Sitting around for half a month
doesn't change that.  Many of the big projects I work on never even see
the light of day because of management cuts, etc.  You get use to it.


#63 of 137 by janc on Thu Dec 4 00:53:15 1997:

The sad thing is that while, yes, in theory there isn't more than a few days
of work to be done to get Grex switched over, in practice it is likely to be
a while before the appropriate people find time to do that work.  When is
unpredictable.  Yes, we could be up on the 670 three days from now, but in
real life there is going to be some random sized interval before that happens.
I worry that people will see the election as an excuse to delay doing work
now, and thus lose what momentum we have.  But most likely we'll use the
December 12 staff meeting as a rabble-rousing session and see if we can get
things moving again then.  "A New Grex for the New Year!"  "Rah, Rah, Rah!"


#64 of 137 by remmers on Thu Dec 4 14:00:39 1997:

Is this primarily work that needs to be done on the 670 before
the actual switch-over?


#65 of 137 by valerie on Thu Dec 4 16:47:58 1997:

Yes.


#66 of 137 by krj on Thu Dec 4 21:58:27 1997:

Can't we set the end of the election as the target date for the cutover?
(Oooh, just what dpc likes, a deadline!  :)  )
 
If the cutover goes sour we can always announce the election results
on M-net.


#67 of 137 by dpc on Fri Dec 5 23:52:01 1997:

Yes, krj, I *do* like your suggestion for a deadline.   8-)
But seriously, I think that if we aim for December 16 we'll probably
get there.
        By the way, how much of a speed increase will we get on the 670?
As srw says, we are *really* hurting on the present machine.  Load averages
routinely over 20 are quite discouraging.


#68 of 137 by kaplan on Sat Dec 6 00:24:09 1997:

Didn't Marcus say he was expecting the 670 to be about 5 times faster?


#69 of 137 by valerie on Sat Dec 6 14:23:44 1997:

I've heard estimates that range from 2 to 5 times faster, though mostly it's
hard to predict.


#70 of 137 by dang on Sat Dec 6 17:13:14 1997:

It's hard to estimate, because the 670 is a multiprocessing machine (2 CPU's)
and we don't know what effect that will have.  Certainly, it will be faster,
but how much faster is a guess.


#71 of 137 by valerie on Sun Dec 7 22:47:39 1997:

Yup, also the 670 has more "contexts" than the current Grex, which means that
it can "remember" what more people are doing before it starts having to write
the information to disk (swapping).  We're not sure exactly how much it will
speed up Grex to have more contexts.


#72 of 137 by srw on Mon Dec 8 02:30:51 1997:

Someone could be doing a security check on the machine right now instead of
waiting until later. 


#73 of 137 by lilmo on Wed Dec 17 02:21:35 1997:

Yea, staff!!!   (Just kind of in general.)


#74 of 137 by dpc on Thu Dec 18 15:56:07 1997:

Valerie says that the staff will try to get the 670 working in
January and not before, because they're worried about what would
happen if we tried to get the thing up and then staff left for
the holidays, followed by a crash.


#75 of 137 by tao on Thu Dec 18 20:42:48 1997:

That's a very reasonable precaution.  New hardware of this
sort shouldn't be set up and then left alone, even for a
week's vacation.  Newly-installed equipment can die despite
staff's best efforts; a component may fail, etc.


#76 of 137 by scott on Fri Dec 19 02:06:45 1997:

Right.  We'd hate to be stuck with stores closed, staffers preferring to spend
time with family rather than in the pumpkin getting it back together.


#77 of 137 by srw on Fri Dec 19 02:23:44 1997:

Actually, I think staff was ready to do this in early December, but the
general opinion was that because of the board election, we should not switch
machines at that time. So we waited until after Dec 15, but by then we are
too close to the holidays. So now we are waiting until January. The machine
is ready to go.  I just received the additional memory for it to bring it to
64M. I got a pretty good deal on it, too, only $248 bottom line for 16 9bit
30pin SIMMs.

Some security checks have been done already, although I would like to see a
few more done. There's nothing to stop that except to get time from the
relevant security experts on the staff.

So right now early January looks pretty realistic.


#78 of 137 by dang on Fri Dec 19 21:38:18 1997:

(That's "bring it to 128M.")


#79 of 137 by srw on Sat Dec 20 21:56:22 1997:

D'oh. 64+64= 128. Yeah. I added 64 to the 64 that were there.


#80 of 137 by tsty on Sat Dec 27 22:56:27 1997:

s'ok, it coulda been 63.99999999M ya know...


#81 of 137 by dpc on Mon Jan 19 16:28:27 1998:

I thought I'd resurrect this item because our load average is around
19.  The last I heard of the 670, Steve Andre had volunteered 
to make some time on his calendar to do the final security checks
needed to start the 670 working.
        Do we know when this might happen?


#82 of 137 by valerie on Tue Jan 20 05:06:14 1998:

19 is pretty low for Grex's load average.  The system is usually pretty
usable into the mid 20's.  However, the dial-ins are insanely lagged since
they moved to the terminal server, so that's probably the problem you're
seeing.

I'm not sure about STeve's calendar.


#83 of 137 by arthurp on Wed Jan 21 04:57:39 1998:

What would be the reason for lag on the modems?  I don't follow.


#84 of 137 by tsty on Wed Jan 21 09:52:40 1998:

maybe i'm not too bright, but... since "the dialins are
insanely lagged since they moved to the terminal server.." *again* ...
WTF put them there?  turn on my light, ok?


#85 of 137 by scott on Wed Jan 21 12:10:48 1998:

The modems are on the terminal server so they can run faster, and so that
there is no ALM (serial card) needed on the 670.  However, there seems to be
some kind of lag effect that started *before* the move to the termial server,
which right now I am thinking is line noise.


#86 of 137 by arthurp on Thu Jan 22 05:09:57 1998:

Line noise related to Ameriwreck work done recently and mentioned in
another item?


#87 of 137 by dpc on Thu Jan 22 21:02:45 1998:

Calling Steve Andre.  Calling Steve Andre.  Come in, Steve!   8-)


#88 of 137 by other on Thu Jan 22 21:27:53 1998:

STeve is busy right now.  Please leave a message at the sound of the tone,
and STeve will get back to you.  Thank you.
                                                        <g>


#89 of 137 by lilmo on Thu Jan 22 22:17:41 1998:

*beep*


#90 of 137 by janc on Sun Jan 25 22:42:19 1998:

Advantages of the terminal server:
  - All modems can be run at 14.4K instead of a few at 9600 and the rest
    at 2400.
  - Reduction of CPU load on Grex from serial I/O interupts.  These were
    especially bad when people zmodemed files to Grex.
  - Possible reduction of flakiness of Grex.  Some of the crashs Grex
    was having a while back may have been related to ALM card problems.
    There are also hints of other problems being caused by bad behavior
    of the ALM card.
  - Elimination of worries about compatibility of ALM card with 4/670.
    The ALM cards are old and dubious technology.  They may be causing
    problems with the newer 4/260 CPU.  They are even more likely to be
    a source of problems with a much newer 4/670 CPU.
  - Reduction of power consumption.  At a guess, we'll probably be using
    $5 to $10 less electricity per month without the ALM card.  This also
    means less heat in the Pumpkin.

Disadvantages of the terminal server:
  - The "please wait a bit" delay.  We may be able to improve this.
  - Some teething problems with lag and lost data that we will almost
    certainly be able to fix, and probably have no relation to the terminal
    server.

Anytime we make any change, there are likely to be short-term problems, some
real, some imaginary.  It's the price of doing business.


#91 of 137 by aruba on Mon Jan 26 03:34:00 1998:

Saving power will be great, if someone will PLEASE MEASURE OUR FREAKIN'
CONSUMPTION, because otherwise we will keep paying what we are now, which is
based on a measurement from over 2 years ago (of the Sun 3).


#92 of 137 by valerie on Tue Jan 27 16:41:29 1998:

Re 91: Just about every month at the monthly staff meeting we talk about this
and someone or other says that they will go measure Grex's power consumption.
It doesn't seem to matter who volunteers to do it; it never does happen.
We definitely need to measure Grex's power usage.  I'm amazed at how long
it's taking.


#93 of 137 by lilmo on Tue Feb 3 22:24:30 1998:

Can a meter be installed?


#94 of 137 by dpc on Wed Feb 4 01:49:03 1998:

Can a Steve Andre-detector be installed?   8-)


#95 of 137 by gibson on Wed Feb 4 07:45:14 1998:

        If someone has an old home electric meter it would be easy to wire it
to plug in the wall and the sun plug into it. Property disposal used to have
some for $5.00 but I haven't been there for some time.


#96 of 137 by scott on Wed Feb 4 12:07:53 1998:

We've got tools to measure power consumption, it's just that it is a real pain
to do (hint:  Grex has to be shut down twice...)


#97 of 137 by gibson on Sat Feb 7 05:51:13 1998:

        You could plug it in when grex is down sometime, you really dont have
to remove the meter afterward unless you wanted to use it elsewhere.


#98 of 137 by rcurl on Sat Feb 7 07:16:04 1998:

The last time STeve measured the power it was with a clamp-on ammeter and
a power cable split so one lead could be clamped around. Easy to split
power cables without turning things off (v..e..r..y carefully...). So,
why not do it without shutting down the system?


#99 of 137 by mdw on Sat Feb 7 09:00:59 1998:

Most of grex's power cables are the round kind with the really tough
outer insulation.  So far as I know, none of the staff has a death wish.


#100 of 137 by scott on Sat Feb 7 13:11:26 1998:

And the power goes thru multiple cables to multiple outlets.  Maybe from the
breaker box, wherever *that* is in the building...


#101 of 137 by valerie on Sat Feb 7 13:38:41 1998:

I've been to Grex's breaker box.  To get there, you need a landlord with a
key, and you need to go through another tenant's space.


#102 of 137 by dpc on Sat Feb 7 16:51:43 1998:

Maybe we could put the clamp-on ammeter across the ears of the 670,
repass the bylays, turn on the power at the breaker box, and see
if the 670 will rise from the dead!   8-)


#103 of 137 by rcurl on Sun Feb 8 05:08:48 1998:

Each cable can be measured separately. It might be useful to prepare some
pigtails - short jumper cables with M/F connectors on ends - with the
wires separated, and put them in the current cables some time the system
is down. Then you can clamp anytime you like. Or better, get an extension
cord that has a cable you can split, which has a multiple outlet junction
box on the end. 




#104 of 137 by gibson on Sun Feb 8 05:32:11 1998:

        I toured the pumpkin today, there's no simple way to do it. There are
too many hook ups.


#105 of 137 by rcurl on Sun Feb 8 06:29:20 1998:

Well, I haven't been there, but that sounds somewhat defeatist.


#106 of 137 by dang on Sun Feb 8 16:11:58 1998:

Oh, it's doable, but not easy.  Rane: we have a quad extension cable with the
wires seperated.  It's just not enough to plug everything, and I wouldn't put
everything on one outlet anyway.


#107 of 137 by rcurl on Sun Feb 8 19:33:04 1998:

So, get another quad ext cable with separated wires?


#108 of 137 by keesan on Sun Feb 8 23:26:32 1998:

We have an electrical meter of the type used residential by Detroit Edison,
and the appropriate socket (don't ask where we got it).  The socket does not
have the cable clamps, they would have to be made.  The meters are sold by
some place in Detroit, don't know where to buy a socket.  Any interest in
this?  We would like it back eventually.  WOrks on 220.


#109 of 137 by scott on Mon Feb 9 00:19:42 1998:

<sigh>  We *have* the tools, just not the discipline to do the work.


#110 of 137 by mta on Mon Feb 9 20:34:12 1998:

According to the MOTD, someone found thetime to take care of this.  Thanks,
folks!


#111 of 137 by dpc on Tue Feb 10 02:28:18 1998:

Unfortunately, we'll have to wait a bit more for Steve Andre to do the
security check on the 670.  At the budget meeting on Sunday, he said
he needed to get some authoritative lists of potential security flaws.
President Jan and others said that we couldn't just fix the most
likely flaws; we have to do the complete list because Grex is attacked
so often by vandals.  He called this necessary wait a pain in the ass.
Which it is.  But we can't risk putting up the 670 until it's done.


#112 of 137 by aruba on Tue Feb 10 04:28:20 1998:

Scott measured the power!  It's about 10 amps.  Thanks Scott!


#113 of 137 by gibson on Tue Feb 10 07:33:36 1998:

        Is that 10 amps 24 h/day?


#114 of 137 by valerie on Tue Feb 10 13:38:20 1998:

DAVE!!!  You were also specifically asked not to discuss that in public.


#115 of 137 by rcurl on Tue Feb 10 19:11:53 1998:

Power doesn't come in amps - that's the current. I've often wondered -
what is the power factor of computer equipment? 


#116 of 137 by mta on Tue Feb 10 19:20:12 1998:

Dave, it's extremely unseemly of you to spread confidential information.
I'd have thought better of you.


#117 of 137 by dpc on Tue Feb 10 20:10:09 1998:

The fact that we don't have a list was *not* confidential.  What *was*
confidential I didn't enter (obviously).


#118 of 137 by remmers on Tue Feb 10 20:49:12 1998:

I wasn't at the meeting, but nothing in #111 struck me as
sensitive.


#119 of 137 by jep on Tue Feb 10 21:11:11 1998:

Please don't misconstrue this question.  I do not mean to attack Dave 
Cahill (or anyone else) in any way.

Why does Dave have secure information for Grex?  As far as I know, he 
isn't in any administrative or staff position for Grex, and shouldn't be 
privy to such information, unless he's offering legal services to Grex, 
or operating in some other "need to know" capacity.

Why is a non-staffer in possession of system security information about 
Grex?


#120 of 137 by dang on Tue Feb 10 23:42:14 1998:

Probably because he showed up at the meeting. (I didn't, I don't know what
was discussed.)


#121 of 137 by other on Tue Feb 10 23:55:08 1998:

having attended the meeting, and being neither staff nor board, i can say that
the only information discussed which was of a sensitive nature was not
terribly sensitive in and of itself, but just fell under the category of
information which it really doesn't benefit anyone to publicize.  no big deal,
just common sense stuff.

don't waste your time asking me for any more details, they will not be
forthcoming.


#122 of 137 by gibson on Wed Feb 11 06:23:14 1998:

        Rane, my 486 pulls 6-8 amps, with a monochrome vga, thru an inveter.
with deration for power loss figure 5-6 amps grid. turn on the printer and
it pulls another 6-8.


#123 of 137 by dpc on Wed Feb 11 14:31:28 1998:

Frankly, jep, I didn't think *anything* discussed at the budget meeting
fell into the category of sensitive security stuff.  Besides, as you
well know, if some such things had been discussed I wouldn't have
understood them.    8-)
        I had intended #111 to be just another in my series of responses
saying essentially "no progress on the 670".  A snoresville response.
        I don't think we'll *ever* get the 670 up, so my suggestion
in #0, that we dispose of the 670, looks like an increasingly good 
idea.  And #0 was entered five months ago today!


#124 of 137 by n8nxf on Thu Feb 12 12:28:19 1998:

(Wow! Patric!  That is one heck of an amp-hungry printer and 486
machine you have there.  What voltage was that current measured at?)


#125 of 137 by dpc on Thu Feb 12 14:30:02 1998:

Was anything said at last night's staff meeting about finishing the
security work on the 670?  Or about finding an alternative if Steve
Andre can't do it in a reasonable time?


#126 of 137 by jep on Thu Feb 12 14:42:12 1998:

#119 was a pointed question.  Nothing should be discussed in a public 
meeting which is not public knowledge.  Anything discussed at the last 
Board meeting should be (and probably is) available to anyone who's 
interested.


#127 of 137 by rcurl on Thu Feb 12 19:23:58 1998:

This stimulated me to measure my system. It draws overall about 1.4 amps
(ca. 160 watts). The CPU takes most of that. I can't even measure the
printers when they aren't printing. 


#128 of 137 by mta on Fri Feb 13 02:49:28 1998:

> Nothing should be discussed in a public meeting which is not public >
knowledge.

Good point.  


#129 of 137 by gibson on Fri Feb 13 03:02:02 1998:

        Mine is all old equipment. The Measurments are from 12 v thru a very
efficient inverter. I under stand if you shop the right components you can
get the whole system under 4 amps at 117v. 12v would probably be 5-6 amps
which still beats the old equipment by a lot. Any one wanting low draw systems
should call Air Castle Enterprises ( or computers, iforget which) in
Southfield.
        If anybody has a CPU to upgrade my 486sx please mail me info.


#130 of 137 by n8nxf on Fri Feb 13 12:21:57 1998:

(4 A @ 117 v ~ 468 watts; 6 A @ 12 v ~ 72 watts, therefore inverter
efficiency = 468 W / 72 W ~ 650%...  Dang good!  ;-)


#131 of 137 by gibson on Sat Feb 14 03:41:20 1998:

        Did i mention math is not my strong point? The 12v is measured, i don't
have an ac ammeter yet so my 2 am guestimate @may be a little off.@


#132 of 137 by n8nxf on Mon Feb 16 13:26:20 1998:

(The power supply in my HP 486DX/66 is rated for 70 watts.  The monitor
is about 200 watts.)


#133 of 137 by rcurl on Mon Feb 16 20:38:47 1998:

Is that measured, or only "rated"? 


#134 of 137 by janc on Fri Feb 20 16:09:43 1998:

Has anyone converted the 10amps number reported above into dollars/month?

I'm not sure if I'm doing this right:

  (10 amps) (110 volts) = (1100 watts) = 1.1 kilowatts
  (1.1 kilowatts) (24 hours/day) ( 30 days/month ) = 792 kilowatt-hours/month

Looking at some of my electric bills, it looks like Det Ed charges about
10 cents per kilowatt hour (or a bit less - about 0.095 dollars per kilowatt
hour).

  (792 kilowatt-hours/month) (0.095 dollars/kilowatt-hour) =  $75.24 / month

Maybe I should be using 120 volts instead of 110.  Then it would be $82 per
month.


#135 of 137 by gibson on Fri Feb 20 23:22:37 1998:

        The figures look right. It's a power hog.


#136 of 137 by tsty on Sat Feb 21 08:47:23 1998:

power factor frm a cpu is most likely not too darn large since
sewitching power supplkies are used and they tend not to have
monster transformers with lots of inductance lag (therefore
a bad power factor).
  
  
if the voltage is actually running around 118-120, using 110 for
figuring watts is probably about right - at least close enough
i would guess.
  
i know weh ahve some honking disks on that machine... even so
that 10 amps does seem a bit much.. was that rms amps or peak amps?  


#137 of 137 by dpc on Sat Feb 21 18:07:59 1998:

I'm freezing this item because it's drifted.  Maybe someone should
start a new "power" item.


There are no more items selected.

You have several choices: