|
|
| Author |
Message |
| 25 new of 65 responses total. |
nephi
|
|
response 25 of 65:
| 
Apr 5 05:48 UTC 1995 |
Is it permanent?
|
selena
|
|
response 26 of 65:
|
Apr 5 05:53 UTC 1995 |
Is ANY memory permanent? I forget.. :}
o
|
srw
|
|
response 27 of 65:
|
Apr 5 06:15 UTC 1995 |
No it will go back to 96MB when we return the borrowed 32MB.
|
nephi
|
|
response 28 of 65:
|
Apr 5 07:58 UTC 1995 |
Thanks.
|
ajax
|
|
response 29 of 65:
|
Apr 5 15:58 UTC 1995 |
Is there much difference in swapping or cache hits that's noticable
in log files or anything? This isn't a scientific observation, but
Grex doesn't really "feel" any faster to me since the addition, for a
given number of users on-line. I would have thought that much more
memory would have made a much bigger difference.
|
steve
|
|
response 30 of 65:
|
Apr 5 16:24 UTC 1995 |
We're cartainly swapping a lot less often, I know that. I think I
do in general feel the different, but yes, it was less than I had hoped
for, too.
I think what this points to is the fact that our little CPU is just
completely swamped these days when we have a full house (like 45+ people)
and all the other things going on here. Having only eight context
switches is whats killing us these days.
Moving onto the SPARC will be a wonderful thing. But even the
4/200 board is gonna get swamped. It's just it will be less swappped. ;-)
|
popcorn
|
|
response 31 of 65:
|
Apr 5 16:48 UTC 1995 |
Re 29: Really? Things definitely seem faster to me (at least when
crazy people aren't running recursive "ls" commands at high priority)
since we added the memory. Also, the load average seems lower since
Greg switched mail from background delivery to queued delivery.
|
lilmo
|
|
response 32 of 65:
|
Apr 5 20:51 UTC 1995 |
Re #30: What are"context switches"?
|
scg
|
|
response 33 of 65:
|
Apr 6 00:31 UTC 1995 |
Although it certainly isn't at th emoment, Grex has recently been feeling
a lot faster to me. The load averages have geenerally been lower too, I
think.
|
adbarr
|
|
response 34 of 65:
|
Apr 6 02:12 UTC 1995 |
Sorry, Steve (scg) it is *you* who are getting slower, Grex is a
constant. Hey, it is very natural to have these feelings - talk
to srw - he has had them for a long time, as have I. It is not
confined to Grex, I am sorry to say. Everything starts speeding
up rapidly once you past 12 - some wierd law of physics I guess.
|
nephi
|
|
response 35 of 65:
|
Apr 6 04:07 UTC 1995 |
This is *drift*, but I think the load averages are getting
higher rather than lower. I have uptime in my .login, and
since the mail tweak or the new dialins, I've often seen
load averages in the twenties through the forties. Before
the changes they were usually in the teens.
How's the Spark coming?
|
popcorn
|
|
response 36 of 65:
|
Apr 6 13:19 UTC 1995 |
Actually, the load average doesn't totally correlate to the system's
response time. STeve and Greg were theorizing the other day that with
more memory, Grex doesn't need to do as much swapping to disk, so at any
given time there are more processes that are "ready to run" instead
"being written to or read from disk". Swapping to disk is slow. The
load average is a reflection of the number of processes in the "ready
to run" state. So, even though the load average is higher, it's actually
a sign that less swapping is going on, which means the system is *faster*.
It might make sense to compare how fast the system *feels*, rather than
using the load average as a yardstick. I wish there were a better way
to quantify the system response time.
|
jep
|
|
response 37 of 65:
|
Apr 7 05:35 UTC 1995 |
With a load average of 15.xx right now, despite my ultra-slow
Internet connection from M-Net through the entire rest of the world
(including a Dixie-Cup and string network in India, I am sure), Grex is
very reasonable to use at this time. No 5 minute delays, no disconnects
from the Internet, a little choppiness, but tolerable.
|
janc
|
|
response 38 of 65:
|
Apr 7 05:49 UTC 1995 |
Moving onto a Sparc? How many years obsolete would that be?
|
gregc
|
|
response 39 of 65:
|
Apr 7 06:35 UTC 1995 |
About 5 years Jan. The 4/200 came out in 89 or 90.
|
adbarr
|
|
response 40 of 65:
|
Apr 7 10:15 UTC 1995 |
There has been some talk (early days of "bad ol' HVFN) about Sun Sparc
20's being used for the project, now HVCN. I believe I saw
a reference to two being used either at Genesee or Detroit F-N. Would
someone comment on capability - desireability of these - ? Thanks.
|
steve
|
|
response 41 of 65:
|
Apr 7 13:03 UTC 1995 |
Yum. Slurp. Drool. Pant. Lust.
|
selena
|
|
response 42 of 65:
|
Apr 7 14:54 UTC 1995 |
Wait- what happened to the mail??
All I know is my friend sent me some mail last night, at 10:30 pm, and
it didn't hit my box 'till 2:30am, and it was sent FROM grex, TO grex!
|
gregc
|
|
response 43 of 65:
|
Apr 7 15:42 UTC 1995 |
When the system load goes above a certain level, the mail system stops
processing mail until the load goes back down.
|
steve
|
|
response 44 of 65:
|
Apr 7 16:26 UTC 1995 |
As things are currently defined, sendmail will not accept mail
from the outside world when the load average goes above 12.
Sendmail will not process mail already in the queue if the load
average is above 20. This figure used to be 8.
|
adbarr
|
|
response 45 of 65:
|
Apr 7 17:44 UTC 1995 |
re 41 - Hmm. Any idea of cost with a UM discount? Are there discounts
for two or three? Always like redundant systems. I suppose
a big cost factor is total RAM. Just think'n.
|
steve
|
|
response 46 of 65:
|
Apr 7 17:57 UTC 1995 |
Yes, RAM is the single biggest expense on a large system
these days, with disk prices falling about every two weeks.
I'll bet (off the top 'o my head) that you could get a 30%
discount if you were able to use UM pricing (can you really?)
But you might still be looking at $8000 - $12000 or something
in that range. If I'm too high in those prices, good. I'd like
to be wrong by a factor of two, at least. ;-)
|
lilmo
|
|
response 47 of 65:
|
Apr 7 18:44 UTC 1995 |
Re #37: *laugh* That's a good one, jep! :-)
Re this item: So, staff, anything new on the to-do list?
|
jep
|
|
response 48 of 65:
|
Apr 8 02:15 UTC 1995 |
(Dixie-Net is my own invention. If you're a marketing whiz, please
contact me with a business proposal.)
|
adbarr
|
|
response 49 of 65:
|
Apr 8 11:01 UTC 1995 |
STeve (#46) - Over the past few months we (HVCN Board) have been
told (nothing in writing) by "highly placed sources, proven to be
reliable in the past") that major discounts enjoyed by certain
humongous educational establishments with large buying power, would
be available to HVCN. I was also informed that NPTN made major
discounts on Sun equipment availble to their affiliates, but those
may no longer be available. All this is merely talk, but the
educational discounts, I believe are there for the asking.
Have also been trying to cultivate some people in Cabletron with
idea of trading discounts for us in consideration of some
warm-fuzzie publicity and good feelings for them. They have
not given specifics but they are still nibbling the hook. They
make a lot of stuff we might use someday, I think. And, they
have made some very tentative statements about considering to
think about the possibility of considering, maybe, additional
discounts beyond the educational discount. Nothing definite,
no promises, but worth pursuing. There is a lot more out there
if we can get to the right people and ask.
Now that I remember, I have heard prices for the HVCN machine n
[delete n] right in the area you are talking about. Am not
sure if those prices are before or after discount. This is
just me talking - but I wonder if HVCN could reciprocate
to Grex by helping fund some major improvements (aside from
the grant stuff) to your equipment as we go along. You have
given much value to us, we should, in my opinion, be prepare
d to return the favor using our skills and contacts. Of course
we need to get our own system established, but I see no
conflict in trying to get two machines for a larger discount,
and thereby gain mutual increased capability. There are
so many ways we could cooperate and both get something better,
while still being distinct and independent. I do not want
to raise hopes, but we should all be thinking about
possibilites here. It will not ever happen, if we do not
maintain open minds about the concepts.
|