[Info-vax] OT: news from the trenches (re: Solaris)

[JF Mezei]

On 15-03-11 16:42, Phillip Helbig (undress to reply) wrote:

> Any idea what the least expensive new VSI-supported x86 VMS system will
> cost (just the hardware)?  Any idea what the least expensive
> non-supported system which will nevertheless run VMS will cost?

I believe that VMware as a host for VMS is on the early wish list for
VMS. This would mean a fairly wide variety of boxes.

For native boot, higher priced servers are more stable in terms of
technilogy/interfaces. This means that decisions on support today will
still be valid a few years from now when VMS on '86 comes out.

Low cost hardware changes like the weather, and decisions made today are
likely to be outdated for low cost residential hardware 2 years from now.

Even for disks, consider SSDs. If you're gonna support them properly,
you want TRIM support with the OS telling the disk when a block is free
so it need not be remapped/copied over when bocks are virtually
re-assigned as part of SSD management.

And if you're gotta do DELETE *.*;*,  do you want VMS to constantly
rewrite to SSD the .DIR file as well as the relevant blocks in
INDEXF.SYS after each file has been deleted ? Or queue all the changed
in memory and ask the SSD to update the changed blcoks once at the end ?

SSDs cannot update a block with updated data.  It will read a series of
blocks (a page), perform the update in memory, then rewrite the page
elsewhere in a free page (one that has been zapped to all zeros so it
can be written to).

TRIM tells the disk file blocks that are now free so they can be zapped
and be made available for writes. But it doesn't do anything for files
such as INDEXF.SYS that are constantly updated in-situ.