Path: utzoo!attcan!uunet!seismo!sundc!pitstop!sun!amdcad!ames!pasteur!agate!garnet!weemba
From: weemba@garnet.berkeley.edu (Obnoxious Math Grad Student)
Newsgroups: news.sysadmin
Subject: Re: "Morris did it"--the new excuse?
Message-ID: <17088@agate.BERKELEY.EDU>
Date: 16 Nov 88 12:20:47 GMT
References: <978@hub.ucsb.edu> <16965@agate.BERKELEY.EDU> <1409@kuhub.cc.ukans.edu>
Sender: usenet@agate.BERKELEY.EDU
Reply-To: weemba@garnet.berkeley.edu (Obnoxious Math Grad Student)
Organization: Brahms Gang Posting Central
Lines: 43
In-reply-to: sloane@kuhub.cc.ukans.edu (Bob Sloane)

In article <1409@kuhub.cc.ukans.edu>, sloane@kuhub (Bob Sloane) writes:
>> Again, what's your point?  From the user's point of view, it's always
>> one reason or another why the computers/networks are not available a
>> certain XX% of the time.  Every time they go down, do your users hunt
>> around for "whom" to blame?

>Yes, at some sites they do.  In our university environment I have heard
>users complain because we take the machine down once a week to perform
>file system backups.  You have to remember that computer centers live
>and work in a political envrionment, as well as a technical one. I
>don't know about the systems you use, but here, every CPU cycle is
>precious, and there are several different users competing for it.  If
>they think even ONE cycle is wasted they will bring it to our attention.

Of course, if anyone points out to any of these users what shitty pro-
grammers they are, and what a waste of CPU cycles their programs really
are--and yes, a lot of genuine nonsense goes on in the world of research
--well, hell, that's too bad.  As you said: politics.

The record of computer inefficiency that I know of was a lattice gauge
theory calculation that took 1000 hours of Cray time.  These physicists
know their physics, but when it comes to computers--just forget it.  I
heard about this from my officemate, who was asked to help figure out
how to perhaps speed things up for the next level.  (No way were they
going to get 10000 hours.  They were all geared up to designing and
building a special purposed device, but they tried seeing if they could
eke some more out with existing hardware.)  He did all the usual.  My
friend noticed numerous symmetries in the problem, he noticed redundant
DO loop recalculations, he noticed tricks with packing very small integer
calculations.  The result: the new version of the same calculation took
one minute.  He did not even need to use assembly and hand vectorization
techniques.  [Now this was something that deserved a Proxmire award.]

Now the above is an extreme, but it illustrates a very very common at-
titude among researchers when it comes to computers: "why think?".  I
have seen it over and over and over again.  CPU cycles are wasted with
abandon in the name of research.

So I'm usually unimpressed with people who complain about lost CPU time
due to downtime.  Not that knowing the truth would ever help when it comes
to harassed sysadmins dealing with annoyed/annoying users.

ucbvax!garnet!weemba	Matthew P Wiener/Brahms Gang/Berkeley CA 94720