Not true!  The TUNE settings were revamped when MPE/XL came along, and again
when the dispatcher was tweaked at about release 2.0.  I also disagree with
your contention that the default settings are "about as bad as they can be".
An expert can often adjust them for certain situations for a positive
effect, but of settings that work in the widest variety of situations you
can't do better.

-----Original Message-----
From: RJ Keefer [mailto:[log in to unmask]]
Sent: Friday, March 09, 2001 11:40 AM
To: [log in to unmask]
Subject: Re: To decay or not to decay - that is the question.....


I beg to differ.  The default settings were developed in 1973 on the Series
II and have never been changed.  they were okay on that machine.  They were
designed for a machine that was terribly I/O bound and constantly waiting
for data from disk.  The machines today and rarely, if ever, I/O bound.
They are almost always CPU or Memory starved.  The default TUNE settings
are about as bad as they can be for today's machines.

Randy Keefer

On Fri, 9 Mar 2001 11:17:04 -0800, Gavin Scott <[log in to unmask]> wrote:

>Paul writes:
>> We currently have out job queues set to decay but are considering
>> modifying some/all to oscillate.
>
>If you have a string on a violin (or similar instrument) which is properly
>tuned, then there are an infinite number of adjustments you can make, none
>of which will be an improvement.
>
>Even if your string is determined to be out of tune, you may be stuck with
>the fact that the rest of the orchestra has "tuned up" to match your
>out-of-tune string, so "fixing" it may simply introduce a more troublesome
>dissonance between the parts of the whole, and it may be better to stick
>with what you have.
>
>By all means play with the :TUNE command if you like, but keep in mind that
>the default settings are what MPE is developed and tested against, and the
>same is probably true for most applications that run on MPE.
>
>More than once have we solved all of a customers horrible performance
>problems by doing nothing more than putting the :TUNE parameters back to
>their defaults.
>
>Science tells us that there's no point in even speculating about things
>which we cannot measure, and playing with the :TUNE command is difficult
>because it is so very hard (in most cases) to determine just *what* the
>results of a change are.  Unless you can scientifically measure the effects
>of each tweak with valid reproducible experiments, it's easy to convince
>yourself that you've made things better when in fact they have actually
>gotten worse overall.
>
>G.