HP3000-L Archives

November 2001, Week 4

HP3000-L@RAVEN.UTC.EDU

Options: Use Monospaced Font
Show Text Part by Default
Show All Mail Headers

Message: [<< First] [< Prev] [Next >] [Last >>]
Topic: [<< First] [< Prev] [Next >] [Last >>]
Author: [<< First] [< Prev] [Next >] [Last >>]

Print Reply
Subject:
From:
Denys Beauchemin <[log in to unmask]>
Reply To:
Date:
Fri, 23 Nov 2001 09:38:05 -0600
Content-Type:
text/plain
Parts/Attachments:
text/plain (228 lines)
Roy obfuscates the issue further.  :)

Let me further explain.  MPE does very little compared to a PC or Mac.  The
MPE OS is minuscule compared to either desktop/laptop OS.  I listed many
things that a desktop/laptop can do which an MPE server can't and should,
never do.

Further a desktop is not built to handle multiple requests simultaneously.
IDE drives are controlled by the motherboard and the CPU.  If you want to
try something, just look at how much CPU is required to read an ATAPI
CD-ROM.  If you want to try something else, launch two disk intensive
operations at the same time.  You will see they are fighting each other over
the drive.  IDE is useless for simultaneous, interleaving access.  That is
why the drives are so cheap.  Now SCSI on the other hand, will do this
nicely and all day long.  But that is why you need a SCSI controller; this
controller offloads all disk and CD-ROM activity from the CPU.
Desktops/laptops do not have SCSI devices (usually) whereas MPE systems only
run with SCSI devices. (I know, HP-IB also, but it's the same.)

You can have the fastest CPU and memory subsystem, but if you can't get to
the disk, it's worthless in an OLTP environment with lots of users.  So yes,
a desktop has the CPU power to do what MPE does, it just doesn't have the
disk subsystem to do it.

Enter SCSI-based servers.  This is where I made the rapprochement with the
A-class system.  Also, the N-class system is very similar and demonstrates
excellent results, which I shall be posting in a separate message about
backup speeds later.

Now to answer a few more points.  The keyboard is still interrupting the CPU
every time a key is pressed.  The mouse is still interrupting the CPU every
time it is moved.  It also consumes CPU.  Don't believe me, get into task
manager and shake the mouse around.  Look at Explorer CPU usage.

The graphics card does indeed offload a lot of work from the CPU, but the
CPU must still feed it constantly as you do things on the system.  I am
reminded of a funny story about someone who didn't understand this concept
either.  I walked into a server room with Proliant servers running Windows
NT all over the room.  I was also amazed to see that all these servers were
running pretty elaborate screen savers.  The OpenGL kind.  I asked the
customer why they did that, (s)he responded they wanted to make sure there
was no screen burn-in.  I laughed and explained that the screen savers where
consuming a lot of CPU, thereby affecting the background processes.  If they
really wanted to avoid screen burn-in, they should just turn off the
monitors or place them in power-save mode automatically.  (S)he thought that
was a great idea and proceeded to turn off the monitors saying (s)he would
look at the power-save mode later.  Of course, (s)he neglected to disable
the screen saver before turning off the monitors...

MPE is designed from the ground up to do OLTP, with batch jobs thrown-in and
to interact with the users in command line fashion.  The VPlus screens have
the added benefit of acting like batch user input, the 3000 never sees user
interaction on the terminal/PC until the enter key is pressed.  This
offloads the CPU further.  The only other thing MPE is designed to do is
communicate with other computers (PC or server.)  This paucity of features
is both a strength and a weakness of the OS.  It is a strength because MPE
is not loaded down with all sorts of junk and thus is able to run really
well with limited resources.  Take IMAGE for example, it is a very basic,
simple database design.  It runs really fast, but you have to code
everything specifically for it.  IMAGE for instance, (pardon the pun) does
not enforce any data checking.  You can stuff any pattern of bits you want
in any type field.  You can have minimal locking, only using this feature to
satisfy the very low requirements of IMAGE itself.  This can make for very
fast applications.

MS-SQL and other RDBMS on the other hand, check everything that you do and
will not let you put even an invalid date in a date field.  All that
built-in checking requires lots of CPU.

Now let's talk about your 8 MHz 286 Vectra where you compare it to your
current 800MHz machine.

The Vectra did not run Windows 3.x, 9x, Me, NT, 2000 or XP.  In fact, unless
you were one of the few who bought Windows 286, it did not run Windows at
all.  Windows (or MacOS) or any GUI is very heavy on the CPU.  The Linux
crowd always guffaws they can run rings around Windows NT on minimal
hardware.  Until they fire up a GUI that is.

There is an old 486-50MHz 24MB (640x480x8) laptop in the office.  It runs
Windows 95.  I just tried starting MS-Word 97 on it.  It takes 24 seconds to
do that.  On my PentiumIII-500MHz 576MB (1600x1200x32) laptop, it takes less
than a second.  When I shake the mouse on the 486, the CPU usage goes up to
8%.  When I do the same on the PIII, it barely registers.  Yesterday, I
listed all the things I can do on the laptop (which is slower than your
desktop BTW.)  You could never attempt that on the 486, and none of it is
possible on the Vectra you mentioned.  The DVD movie alone requires a lot of
power, far more than even the 928 has.  The constant decoding of the flow of
data (1.5MB/second or thereabouts) and subsequent rendering on the
monitor/panel is much more than the 928 could ever do.

So, I say again, stop comparing an MPE server designed for command line or
batch interaction for multiple users to a multimedia desktop or laptop. It's
a waste of time..., but it's fun, I know.  I am just glad that MPE managed
to acquire some of the hardware technology of the PC before then end.  By
this I mean of course, the PCI architecture.

Kind regards,

Denys. . .

Denys Beauchemin
HICOMP
(800) 323-8863  (281) 288-7438         Fax: (281) 288-7438
denys at hicomp.com                             www.hicomp.com

-----Original Message-----
From: HP-3000 Systems Discussion [mailto:[log in to unmask]]On Behalf Of
Roy Brown
Sent: Thursday, November 22, 2001 8:16 PM
To: [log in to unmask]
Subject: Re: Contingency Planning

In message <[log in to unmask]>, Denys Beauchemin
<[log in to unmask]> writes
>Comparing a desktop or laptop computer to a 918 or 937 is an exercise in
>absurdity.

But as, in a few years time, there won't be any 918s or 937s around (or
at least, nothing to be their contemporary equivalents), we have to
start thinking what else might do the job.

> First of all, the MPE system does very little compared to a Mac
>or a PC desktop or notebook.

So it ought to be easy for the desktop to carry the workload.

>Second, most anyone with an inkling of performance knowledge would
>laugh at anyone trying to use a desktop as a server.   Your 918 has one
>or more SCSI disk drives whereas the vast majority of desktops and
>laptop have a single IDE disk drives.  (Ok, my laptop has 3 disk
>drives, but I'm a weirdo and they are all IDE, not SCSI.) Third, the OS
>of a Mac or a PC is so not geared for multiple users and batch jobs,
>unless you use Windows NT/2000/XP.

1 IDE here, pretending to be two....

So it's hard for the desktop to carry the workload.
>
>But let's see if we can simulate an MPE environment on a desktop.  What
>would be needed?  Ric mentions 30-45 developers on a 937.  Grant me a PCI
>desktop with a network card, one or two disk SCSI disk drives and Windows
>2000 Server.
>
>Ok, the first thing is to disconnect the mouse, monitor and keyboard from
>the desktop.  The HP 3000 does not have a mouse or a monitor or a keyboard,
>no reason to burden the desktop with those things either.  Besides, they
are
>what consume most of the CPU on these systems.

My mouse and keyboard don't go any faster than they did on my 8mHz
Vectra 286, and I don't think that (like a Winmodem say) their
functionality has been offloaded onto the CPU any.

My current machine has 800mHz under the hood. 100 times more.

WRT the graphics, I'm running 1280*1024*32m colours (24 bits, call it
32), whereas before I was running 640*480*256 colours (8 bits). I get
factor 17. But my CPU is 100 times faster. And oops, I forgot the
graphics card with its hardware assist. Actually, I think that alone is
several orders of magnitude faster than the whole Vectra was.

So if you are *serious* about this assertion, better tell me what I've
missed in this logic.

>Next, you only access the desktop via the network. Just like an MPE system.
>Ok, now connect a bunch of people over the network and have them type their
>programs, line by line into an editor, no GUIs allowed.  When they are
>finished, they can compile their programs, again no GUIs allowed, simple
>command line stuff.  BTW, you can run some batch jobs in the background, no
>GUIs.
>
>See it's running well, no problems.  BTW, the system I described above is
>very similar to an A-class box, don't you think?

It's not very similar to what people *want* though, is it?

No good having developers unless people are going to run their programs
to assist in the actual functioning of their business.

>Get real and stop with the silly, pointless comparisons.

What you mean like talking about developer workloads, when the real
issue is production ones?

>  A desktop or laptop has many uses, an MPE box only has one, to run
>command line programs really, really fast.

The people I support, who spend their days in a menu-driven VPLUS
system, would find that a *very* odd assertion. Not only because the
command-line interface is long gone (VPLUS was an advanced GUI in its
day, no matter that it hasn't kept up).

But because, although speed is good, they reckon that just as long as
they have enough power to keep up with their typing, pretty much, they
would prefer the MPE box to give them a *balanced* workload. And of
course, run batch jobs too, as fast as possible in the gaps. And print
stuff. And of course, not break...

> You will never play music, burn CDs, play video games,
>work on digital photographs, scan in documents, watch a DVD movie, do
Visual
>BASIC or Visual C++ development, browse the web and dictate documents on
the
>HP 3000.  I can do all that, at the same time on my notebook PC.  (Well
most
>of it, it's kind of silly to watch a movie, listen to CDs and dictate a
>document all at the same time.)

I'm still totally lost in what you are saying. Is this supposed to be
good or bad? Because a PC *can* do these things, does that mean it *has*
to? Does that mean it *can't* do anything else?

I'm not trying to have MPE run like a PC. I'm thinking of a PC running
like MPE. The whole of the above, maybe, can be summarised by my
observation that sport car (PC) <> panel van (HP3000 running MPE); even
though the power of the engines may be about the same, the way they are
harnessed and the load capacity (1 person extremely fast, but not much
luggage, or a *load* of luggage quite quickly) differentiate them.
--
Roy Brown        'Have nothing in your houses that you do not know to be
Kelmscott Ltd     useful, or believe to be beautiful'  William Morris

* To join/leave the list, search archives, change list settings, *
* etc., please visit http://raven.utc.edu/archives/hp3000-l.html *

* To join/leave the list, search archives, change list settings, *
* etc., please visit http://raven.utc.edu/archives/hp3000-l.html *

ATOM RSS1 RSS2