[mod.computers.vax] VMS: questions about LP11s and file

McGuire_Ed@GRINNELL.MAILNET (08/11/86)

>Date: 8 Aug 86 08:15:00 PDT
>From: "Oberman, Kevin" <oberman@lll-icdc>
>Subject: RE: VMS: questions about LP11s and file secu
>
>I have a possible (and maybe reasonable) solution to the problem of cluster
>disk security from users in a certain group. (i.e. Students).
>
>The students would be given an identifier (e.g. STUDENT). The MFD on the
>disks in question could then be given an ACL prohibiting access by a holder
>of the STUDENT identifier. I don't believe there is any way to access the
>disk if you can't access the MFD.

>A possible ACE for the MFD would be:
>$ SET DIRECTORY/ACL=(ID=STUDENT,ACCESS=NONE) disk$name:[000000]

I infer that there would be an extra ACL test for every attempted file
access for that disk.  Has anybody got any good feel for the affect that
this solution would have on performance?  The documentation warns that
ACLs can have a real impact.

An additional note about the performance of LP11s on an 8600 Unibus. One
DEC source who is in a position to know says that LP11s work fine on a
Unibus.  However, Our Field Service engineer maintains that any non-DMA
device will substantially degrade performance of the 8600, having been
warned by the instructor at a seminar he attended.

The theory is that BR-style Unibus transfers interact with the pipeline
instruction processor of the 8600.  Because BR-style transfers must
occur between instructions, the pipeline has to be emptied.  Since the
pipeline is often empty, fewer instructions can be executed per real-
time-unit while files are being printed.  We are probably going to try
it and see what happens.

There is going to be a session at DECUS on 8600 Unibus performance,
which should address the issue of non-DMA devices.