Article 125183 of comp.os.vms:
In article <5mlc9e$ah6@tooting.netapp.com>, guy@netapp.com (Guy Harris) writes:
> Glenn C. Everhart <Everhart@GCE.Com> wrote:
>>The key to Galaxies is performance a factor of 10 (conservatively!) better
>>than ANY other OS, on the same iron, gained by using VMS instead of
>>some other OS.
> 
> So what about VMS - as opposed to, say, Windows NT, or OS/390, or... -
> gets you the factor of 10?  (This is not a rhetorical question, nor is
> it a troll.)

	Looking in Appendix H (what a wonderful coincidence!) of an old 
	Internals and Data Structures we read:

	"RMS [VMS's Record Management Services] transfers a bucket
	of data on a process's behalf from a file into a buffer in
	memory.  An RMS local buffer is mapped in process space and is
	available to only one process.  A global buffer is mapped in
	system space within a VMS global section and can be shared by
	any process on the system.  Global buffers, however, cannot
	be shared by process on different VAX/[Alpha]cluster nodes."

	Galaxy will change all that.  You know, I gotta get used to
	saying Galaxy.  I blame Terry S., he writes "Galaxies".  Guess
	Galaxies would be more than one.  Could this be
	done on NT or OS/390?   How much memory is directly addressable
	in OS/390?  Seems the mainframe folks are taking a different
	approach and Sun is copying mainframes somewhat.  Search 
	http://www.dejanews.com/ for: Andrew Harrison and COMA

	Seems COMA is Sun's future.  Local domains in the E10000 copying
	from a memory pool to attraction memory.  To gain background
	on COMA see:

http://playground.sun.com/pub/S3.mp/simple-coma/isca-94/paper.html

	As I pointed out to Andrew, why copy?  Why not use it?  Poses
	a problem, how do you lock it down across OS's?  Do you need
	an RMS layer?  Sure would help, eh?  Here is more detail on Galaxy:

http://www.openvms.digital.com/affinity/presentations/wave4/sld043.htm
http://www.openvms.digital.com/affinity/presentations/wave4/sld044.htm

	From slide 44:

"OpenVMS can partition the system using the Galaxy Software Architecture which
is a model for Adaptive Partitioned Multiprocessing (APMP).  It defines how
multiple instances of operating systems execute cooperatively in a single
computer.  OpenVMS APMP can load balance the resources among the partitions, in
real-time, (e.g., hot swapping CPU's, memory, disks, etc. with no need to
reboot) to effeciently allocate them where most needed."

	CPUs leaving and joining the Clusters.  My, my, my . . .

	One can imagine that VMS engineering is pumped up.  When will
	NT get this?  Well... might be about 5 years.  After all, they
	are a bit behind now in some deliverables and 64-bit memory
	(sorta) is betaing the beginning of next year.

	VMS engineering delivered 64-bit one full year ahead of schedule.
	They can move very quickly.  Especially when excited.  Excitement
	comes natural when you are TOTALLY re-defining what a high-end
	solution is all about. . . 

	Run with the Best, VMS!

		Zing!!!

				Rob