[Info-vax] INIT/HIGHWATER

[Steven Schweda]

On Feb 5, 7:57 am, hel... at astro.multiCLOTHESvax.de (Phillip Helbig---
undress to reply) wrote:
> How much of a performance hit is /HIGHWATER?

   (I tried to post this days ago, but Google Groups now seems
to need JavaScript enabled before it will post anything, and
it took me a while to guess that.)

   It depends.  If your software always writes files from
beginning to end, and if it sets the sequential-access-only
(SQO) flag, then highwater marking may make little if any
difference.  If your software dances around in a file while
writing to it, or if it fails to set the SQO flag, then you
may get a significant slow-down.  (And, with highwater marking
and without SQO, allocating large chunks of disk space can
lock up a disk for many seconds, or minutes.)

   With a little fooling around, UnZip can provide a
reasonable test case.  Look for "sqo" in [.vms]vms.c, and see
what happens when extracting a large file, say, 2GB or larger,
if you disable the code where the SQO bit is set.

   Until large-file support was added to Zip+UnZip, I heard no
one complain about the awful performance.  When I started
testing the stuff with large files, I found it unbearable, so,
once someone explained how to solve the problem, I put in the
SQO stuff about as fast as I could.

> Obviously, very vague questions and the answer is "it depends", but I'm
> looking for ballpark figures.

   It depends too much to allow even a rough estimate.