The question is about memory managment performance, In the
manuals an acceptable number of pagefaults for OpenVMS is
defined as an average of 100 pagefaults for each VUP on the
system. Even on OpenVMS 6,2 and on Alpha systems this is not
changed since the good-old VAX 11/780.
In my opinion I can't use the formule 100 * Vup-rate for the current
high performance VAX & Alpha systemen. The same counts for
the percentage of hardpagefaults.
Which number can I use as a rule-of-trumb for the pagefault
rating on the current VAX and Alpha models
Thanks,.
The current "thinking" on page faulting is quite simple: understand
what your paging characteristics are and, where possible if the
faulting can be reduced by increasing working set sizes, do so.
If a number is really desired, start out with somewhere between 500 and
1000 faults as a point where you begin investigating whether page
faulting can be reduced. Once you have adjusted the system as well as
possible to avoid faulting, use it's steady state faulting rate as your
new threshold.
Please note that hard faults, those faults that require reading the
page(s) from disk, should still be reduced and avoided whenever
possible.
