RE: subject: iSeries Legends

On Tue, 24 Jun 2003, Joe Pluta wrote:

> It's not about load, it's about complexity of work.  Web/mail serving is
> orders of magnitude simpler than multitasking OLTP.  But you know what?  I
> don't know your machine, and it's your contention that your box works as
> hard as an AS/400.  I think you're absolutely off your rocker, but there's
> no good way to really resolve this.  It could escalate into a flame war, or
> at the least cause bad feelings, and rather than go there, I'm simply going
> to bow out with a "no decision".
>
> However, I will state that it is MY contention that an IDLING AS/400 is
> doing more complex work than a fully loaded dedicated web/mail server.
> That's my contention based on a career of writing applications and operating
> systems.  I know in a general sense what a web server is doing down to the
> interrupt level, and it's nothing compared to a self-adjusting system like
> OS/400.  But I probably can't convince you, so I'll just let my statement
> stand as my personal opinion.

Interesting.  I wonder if from the computer's point of view there is such
a thing as a complex vs. a simple task?  IOW, does the computer know it is
doing something complex vs. something simple?  I doubt it.  Whether it is
dedicating substantial resources to something we perceive as simple or
complex it is all just 1's and 0's to the computer.  So the complexity of
the task makes no difference.

What could make a difference is the different parts of the machine that
are being exercized.  As you use more and more components you become more
and more likely to trip over a bug.  The interesting conclusion of this
statement is that a machine with fewer subsystems should be more stable.
Therefore, if we believe that the iSeries has many subsystems all active
all the time, we should also conclude that it is less stable.

But less stable than what?  Apparently Scott's toaster - and we are
correct in our conclusion.  But does the argument hold when comparing
OS/400 to other OSes?  Too tricky to tell, since we don't have a good way
to compare subsystems.  So we have to throw out that argument and work
largely on empirical evidence.  And the evidence is that the iSeries is
very stable and reliable, as are several other OSes.

I guess the statistic we are looking for is unit work done over uptime.
That would be a really interesting research topic.

However, I think you misspoke when you said that your iSeries does complex
tasks when it is idling.  That is a contradiction in terms.  If it is
actively doing work then it isn't idle.  Furthermore, is it a good thing
that the iSeries is contantly busy even when there are no user tasks to
execute?  I think there is a strong argument that the answer is no.

James Rich