[Info-vax] Raxco VMS Tuning Seminar Notes

[jls]

On Sun, 31 Oct 2010 07:01:13 -0700 (PDT), Neil Rieck
<n.rieck at sympatico.ca> wrote:


>
>I think it is wise to point out that the Raxco Notes were based upon
>1987 VAX systems. In those days, the majority of systems didn't have
>sufficient electronic memory. If I need to support (too) many
>processes using too little memory, then I would always set PFRATL
>above zero. If my VAX was running an industrial process then I
>probably would not.

Again with the "always".  The drawbacks with this memory management
technique are:
	1. it uses CPU to constantly manage memory, thus taking
available CPU from users  This is a trade-off that is not adequately
understood by many people.
	2. it only takes memory from ACTIVE users, not idle users.
This forces the active users to pagefault much more to get work done.

Now, I'm not saying that use of PFRATL is never a good idea, but I
found much better performance management results (depending on the VMS
version) by swapping out entire processes by manipulation of other
params (swpoutpgcnt at the least, but probably others as well).  Note
that this method reclaims memory from IDLE processes so that active
processes do not have to work as hard... and swapping back in is done
with ONE I/O, rather than lots of I/Os for hardfaulting.  

>
>But things changed with Alpha. Many of theses systems were installed
>with much more memory and the autogen script seems to disable
>automatic working set adjustment (at least AWSA reduction)  whenever
>more than 500 MB is present.
>
 In early VMS VAX versions you wanted to avoid swapping as much as
possible, but this was much improved at or around V4.0 or 5.0 (forget
exactly which).  Imho, at that time the "costs" of using PFRATL for
memory management were too great - we didn't buy our systems to manage
memory, but to run our workloads.

I believe I complained to someone way-back-when that the memory
management education need to be modified to accommodate this, but it
never went anywhere.