PL/1 mainframe Multitasking 
Author Message
 PL/1 mainframe Multitasking

Hi
At my work we use:
5668-910 IBM OS PL/I OPTIMIZING COMPILER  VER 2 REL 3 MOD 0    
on a 390 system.

A few years ago when I was learning PL/1, and had little job
responsibilities, I played with multitasking. If I remember I had 2
versions of the same job.

Bottom line: both performed much slower and more expensive than the
equivalent sequencial version of the program. Although it was a simple
program I hoped it would demonstrate the feasibility of multitasking.
Is this the norm or did I just do something stupid?

I still have the source code if anybody's interested, must just sort
out which one did what.

Regards
Leon



Tue, 31 Dec 2002 03:00:00 GMT  
 PL/1 mainframe Multitasking
Leon,
I assume that you mean multitasking in a batch environment.
On a modern mainframe (sorry, enterprise server) there are
usually few benefits in multitasking.
The operating system is multitasking the various programs.
File buffering also reduces its advantages
Multitasking will be more expensive.
The same work has to be done, plus the multitasking overhead.
For it to be of benefit you must be prepared to give more of
a resource you can afford to gain a resource you want.
You could, for instance, be short of elapse (wall clock) time
and be prepared to pay in terms of other resources - cpu etc.
Now if you had to read through an enormous number of records
you could trigger (say ten) tasks each starting at different
record numbers (or keys) and each processing (say one tenth)
of the file.  This will reduce your elapse time.
It will cost cpu, memory, disk contention.
If you will multitask, then because it has reentrant procedures
PLI is the best language to do it in


Quote:

> Hi
> At my work we use:
> 5668-910 IBM OS PL/I OPTIMIZING COMPILER  VER 2 REL 3 MOD 0
> on a 390 system.

> A few years ago when I was learning PL/1, and had little job
> responsibilities, I played with multitasking. If I remember I had 2
> versions of the same job.

> Bottom line: both performed much slower and more expensive than the
> equivalent sequencial version of the program. Although it was a simple
> program I hoped it would demonstrate the feasibility of multitasking.
> Is this the norm or did I just do something stupid?

> I still have the source code if anybody's interested, must just sort
> out which one did what.

> Regards
> Leon

Sent via Deja.com http://www.deja.com/
Before you buy.


Tue, 31 Dec 2002 03:00:00 GMT  
 PL/1 mainframe Multitasking

Quote:

> A few years ago when I was learning PL/1, and had little job
> responsibilities, I played with multitasking. If I remember I had 2
> versions of the same job.

> Bottom line: both performed much slower and more expensive than the
> equivalent sequencial version of the program. Although it was a simple
> program I hoped it would demonstrate the feasibility of multitasking.
> Is this the norm or did I just do something stupid?

This discussion came up a while ago, and you might want to check the
archives (at deja.com, or are these discussions also archived
elsewhere?)

The only time multitasking will perform better is if the program itself
does a lot if independent functions which require waiting.  
For example, before large memory buffers and cached DASD, IO used to
require a fair amount of waiting, even though the OS tried to
overlap it with processing.

Multitasking now is more useful for "server"-type processes which
perform various functions independently for multiple requesters, such as
an internet FTP server.  Servicing each request may take about the same
time or slightly longer as a single-tasked version, but the total number
of requests services and the "responsiveness" will be better with the
multi-tasked version.



Tue, 31 Dec 2002 03:00:00 GMT  
 PL/1 mainframe Multitasking

Quote:

> Hi
> At my work we use:
> 5668-910 IBM OS PL/I OPTIMIZING COMPILER  VER 2 REL 3 MOD 0
> on a 390 system.

> A few years ago when I was learning PL/1, and had little job
> responsibilities, I played with multitasking. If I remember I had 2
> versions of the same job.

> Bottom line: both performed much slower and more expensive than the
> equivalent sequencial version of the program. Although it was a simple
> program I hoped it would demonstrate the feasibility of multitasking.
> Is this the norm or did I just do something stupid?

> I still have the source code if anybody's interested, must just sort
> out which one did what.

> Regards
> Leon

Sooner or later (probably later), you will use the most current compiler
version 'Visual Age PL/I for OS/390' and this version will NOT support
multitasking anymore.

--
Daniel
------------------------------------------------------------
visit us at:
http://www.winterthur.com

Sent via Deja.com http://www.deja.com/
Before you buy.



Tue, 31 Dec 2002 03:00:00 GMT  
 PL/1 mainframe Multitasking

Quote:
>Subject: PL/1 mainframe Multitasking
>From: Leon
>Date: 7/14/00 4:20 AM Central Daylight Time

>Hi
>At my work we use:
>5668-910 IBM OS PL/I OPTIMIZING COMPILER  VER 2 REL 3 MOD 0    
>on a 390 system.

>A few years ago when I was learning PL/1, and had little job
>responsibilities, I played with multitasking. If I remember I had 2
>versions of the same job.

>Bottom line: both performed much slower and more expensive than the
>equivalent sequencial version of the program. Although it was a simple
>program I hoped it would demonstrate the feasibility of multitasking.
>Is this the norm or did I just do something stupid?

>I still have the source code if anybody's interested, must just sort
>out which one did what.

>Regards
>Leon

Not sure if it applies to your case, but a common multi-tasking mistake is to
"parallelize" a number of tasks that try to simultaneously access the same
resource.

For example, at one of my employers, a single job stream with 20 sequential
steps was split into 20 concurrent jobs in an effort to improve throughput.

As you may have guessed, all 20 tried to access the same disk input files,
causing the electromechanical disk drives to thrash relentlessly as the heads
moved all over creation trying to satisfy all 20 jobs, ultimately leading to a
massive degrade in performance over the original "bad" sequential job stream.



Wed, 01 Jan 2003 03:00:00 GMT  
 
 [ 5 post ] 

 Relevant Pages 

1. Multitasking in PL/I

2. PL/1 (370 mainframe) to PC microcomputer code cross compiler

3. PL/1 (370 mainframe) to PC microcomputer code cross compiler

4. PL/I on a Mainframe

5. PL/1 and testing mainframes past year 2000

6. PL/I mainframe manuals wanted

7. Mainframe PL/I

8. \\\\\\\\\**** Mainframe Programmer/Analyst - REXX or PL/1 ****/////////

9. Logo Komeniusz PL (Logo Comenius PL)

10. Derivation of PL/I (was Usenet group for PL/M language)

11. Mapping local files to FILE declarations in PL/I with IBM VisualAge PL./I for Windows

 

 
Powered by phpBB® Forum Software