| T.R | Title | User | Personal
Name | Date | Lines |
|---|
| 206.1 | "treat memory as the boot device" | FROST::PIPER | bill piper | Thu Feb 20 1986 17:14 | 9 |
| It's something like "Use the pseudo-disk instead of the boot device".
After that, all swapping and such is done in memory, where the pseudo
disk lives. I think this flag was invented when we figured that a
MicroVAX could have only the TK50 as a load device.
I expect to see Jay Olson answer this in the MicroVAX conference one
of these days.
-piper
|
| 206.2 | I want that psuedo disk driver. | RICARD::HEIN | Hein van den Heuvel, Valbonne. | Fri Feb 21 1986 04:00 | 12 |
| Hmmm,... Pseudo disks... Often I recieved requests for that. Almost
as often I could advice alternative sollutions, but not always.
I never had enough need to write the pseudo(?)-disk driver but if
such an animal exists then I would like to play with it. Where can
I find it? on the source kit? Could the psuedo disk facility mentioned
by bill be used in a normal vms system?
Eg; Assuming price is not important but performance is. Given a
system with a 10Mb RDB database, could I copy it into memory, use
it there (keeping the journalling on a real disk) with 'the speed
of light' and copy it back to a real disk every now and then?
Hein.
|
| 206.3 | | ULTRA::PRIBORSKY | Tony Priborsky | Fri Feb 21 1986 09:28 | 6 |
| It's PDDRIVER (probably on the �VMS kit only). However, I don't
recommend it for the use you're suggesting. The "virtual disk"
is taken from non-paged pool. Second, the performance is not going
to be good since the driver uses the IOC$MOVFRUSER and IOC$MOVTOUSER
routines to move the data to/from the virtual disk. These routines
operate at FORK IPL, and move a byte at a time.
|
| 206.4 | Other Ideas? | RICARD::HEIN | Hein van den Heuvel, Valbonne. | Fri Feb 21 1986 09:59 | 6 |
| Re .3: Yes, that does not seem an optimal implementation for
the suggested use. It could be a good starting point though!
Any other ideas? Should we start a base note on this topic?
Does anyone remember where he/she has seen this discussion before?
Hein.
|
| 206.5 | BULOVA::VWS discussed PDDRIVER | FROST::PIPER | bill piper | Fri Feb 21 1986 11:30 | 6 |
| re: .4
This was discussed in the VAXstation notesfile (now conference). That's
BULOVA::VWS (hit KP7). Don't remember the number.
-piper
|
| 206.6 | Another reference | AURORA::HALLYB | | Tue Feb 25 1986 11:31 | 6 |
| It's also discussed in the VMS Performance database, AURORA's VTX
server. As it turns out, using PDs is a big win for one- and two-block
reads. The tradeoff is more CPU for more speed. There are places
where this can be worth doing, but it's not a panacea.
John
|