[Info-vax] OpenVMS I64 V8.1 "Evaluation Release"?

[Paul Sture]

On Fri, 23 Mar 2012 07:25:21 -0700, John Wallace wrote:

> Denial of service through shared resource exhaustion, or the avoidance
> thereof, is part of why VMS has lots of separately manageable quotas,
> and a specific privilege to allow quota checks to be overridden. 

If you are thinking of EXQUOTA there, that's for disk quotas only.

> VMS sysadmins can largely ignore these quotas if they wish, and often do
> without much adverse effect. But where it matters, sysadmins have the
> option to manage quotas securely if they so desire, an option which few
> other OSes offer.

The negative side to that is that every now and again, quotas which are 
too low get implemented.  For either V8.2 or 8.3 the recommended default 
process quotas were given a bump, but unless you read and implemented the 
changes recommended in the release notes, you would carry on using older 
recommended values.  The SYSTEM and DEFAULT UAF entries were mentioned 
specifically, but of course where you have existing users created with 
previous DEFAULT values, you needed to address those as well.

The classic case where the DEFAULT values were overlooked was in 
migrating from VAX to Alpha where SYSUAF.DAT was simply copied over. I 
assume a similar case exists with migration to Itanium too.

> The downside of this is that some operations on VMS incur a quota-
> accounting (and privilege checking) overhead which slows them down
> slightly relative to the equivalent operation on a no-quota OS which
> doesn't have fine-grained privileges. But there is no such thing as a
> free lunch.

No such thing as a free lunch indeed, but when I ran the same large data 
conversions on an Alpha and on a Mac I found that the Alpha elapsed times 
were something like a third of the Mac's, despite using the same CPU time 
on both platforms.  I put that down to a better scheduler on VMS, though 
that's probably not the whole story.  In other words, swings and 
roundabouts apply.

-- 
Paul Sture