[Info-vax] OT: Steve Wozniak

[Neil Rieck]

On Aug 24, 1:49 pm, Keith Parris <keithparris_nos... at yahoo.com> wrote:
> Neil Rieck wrote:
> > Our databases are based upon RMS-ISAM rather
> > than something SQL-based. This means that a majority of our databases
> > are sitting in the XFC (extended file cache) of main memory because
> > our processes only need 1 GB.
>
> XFC caching is a good thing, but you should probably look into enabling
> RMS global buffers, as they can be even more effective.
>
> Many of the former objections to using RMS global buffers have been
> removed. They used to be scanned sequentially for lookups, but 7.2
> introduced hashed lookups. They used to be limited in numbers by a
> 16-bit signed count word, but as of 8.3 the count is a signed 32-bit
> longword. They used to be limited by P0 address space, but as of 8.3 can
> now be in 64-bit P2 space. And the RMS locking algorithm improved in
> 7.2-1H1 such that lock rates are dramatically decreased when global
> buffers are enabled.
>
> Seehttp://www2.openvms.org/kparris/HPTF2008_Sizing_RMS_Global_Buffers.ppt
> for more info.
>
> Keith

Your PPT contains some intriguing ideas which I have shared with a
dozen people at my place of business.

But something else happened yesterday which I would like to share with
all of you: There currently seems to be a small spike in the volume of
used Alpha hardware. Like always, money is tight but we were finally
able to acquire enough used hardware to build a hot-standby DRP
system. We were testing our production software on the DRP system an
noticed that disk-based I/O was twice as fast as production. At first
we thought it might be something simple like disk fragmentation but
further investigation showed that the DRP HSZ-22 (RA 3000) contained
twice the amount of RAM found in our production controller. Time to
add memory to the production RAID controller (or at least swap them
ASAP)

Neil Rieck
Kitchener/Waterloo/Cambridge,
Ontario, Canada.
http://www3.sympatico.ca/n.rieck/