On 6/21/07, Thomas Charron [EMAIL PROTECTED] wrote:
I truly cannot believe, after all of the off-topic conversations
we've had, how anal retentive the list has become recently.
That should be hyphenated as anal-retentive.
;-)
-- Ben
___
On 6/21/07, Paul Lussier [EMAIL PROTECTED] wrote:
Someone just asked me if I had ever heard of an Intel-based system
with application-accessible non-volatile RAM.
How about a PC with a UPS? ;-)
-- Ben
___
gnhlug-discuss mailing list
On 6/21/07, Jon 'maddog' Hall [EMAIL PROTECTED] wrote:
And there's the retro naming of SunOS 4.x to Solaris 1.x.
ummm, that was way more than a retro naming.
SunOS was based on the BSD kernel and the BSD code, modified a long time
under Sun.
Solaris was based on System V.4, with Sun
On Jun 21, 2007, at 16:38, Paul Lussier wrote:
I don't think they want a drive, they want something like a
battery-backed cache but accessible from user space.
Somebody who's done systems programming more recently than me, please
speak up, but if you do find a suitable device that looks like
I don't think they want a drive, they want something like a
battery-backed cache but accessible from user space.
Somebody who's done systems programming more recently than me,
please speak up, but if you do find a suitable device that
looks like a drive, you might be able to create a big file
Michael ODonnell writes:
I don't think they want a drive, they want something like a
battery-backed cache but accessible from user space.
Somebody who's done systems programming more recently than me,
please speak up, but if you do find a suitable device that
looks like a drive, you
From: Bill McGonigle [EMAIL PROTECTED]
Date: Fri, 22 Jun 2007 10:01:46 -0400
Cc: [EMAIL PROTECTED]
On Jun 21, 2007, at 16:38, Paul Lussier wrote:
I don't think they want a drive, they want something like a
battery-backed cache but accessible from user space.
Somebody who's done
A USB thumb drive might be good for that...
On 6/21/07, Paul Lussier [EMAIL PROTECTED] wrote:
Hi all,
Someone just asked me if I had ever heard of an Intel-based system
with application-accessible non-volatile RAM. The idea is that OS
could move things out of swap and/or system memory
looks like a drive, you might be able to create a big file on
it and mmap(2) it into system memory.
I was thinking along those lines, too, but unless you can
force every write to the memory associated with the mmap'd
device to immediately trigger a (synchronous) flush to that
device the
Michael ODonnell writes:
You could programmatically force
a flush after every write, but if you're willing to do that
then there's probably a lot better approaches than this one...
There are better solutions than this, but if somebody were to want
to implement one, they should do their due
On 6/22/07, [EMAIL PROTECTED] [EMAIL PROTECTED] wrote:
From: Bill McGonigle [EMAIL PROTECTED]
Date: Fri, 22 Jun 2007 10:01:46 -0400
Cc: [EMAIL PROTECTED]
On Jun 21, 2007, at 16:38, Paul Lussier wrote:
I don't think they want a drive, they want something like a
battery-backed cache
On Fri, 2007-06-22 at 09:33 -0400, Tom Buskey wrote:
On 6/21/07, Jon 'maddog' Hall [EMAIL PROTECTED] wrote:
And there's the retro naming of SunOS 4.x to Solaris 1.x.
ummm, that was way more than a retro naming.
SunOS was based on the BSD kernel and the BSD
On 22 Jun 2007 11:21:40 -0400, Kevin D. Clark [EMAIL PROTECTED] wrote:
Michael ODonnell writes:
You could programmatically force
a flush after every write, but if you're willing to do that
then there's probably a lot better approaches than this one...
There are better solutions than this,
Ben Scott [EMAIL PROTECTED] writes:
On 6/21/07, Paul Lussier [EMAIL PROTECTED] wrote:
Someone just asked me if I had ever heard of an Intel-based system
with application-accessible non-volatile RAM.
How about a PC with a UPS? ;-)
This is for a cluster of systems. We don't want to have
Worst case, you can open using O_DIRECT, which specifies direct
read/right access. The only limitation really is the reads/writes
need to be aligned to the block size of the device.
Better would be to tell the OS it's OK to cache it, but a direct
write is needed.
Negative. Even though
So, what happened at Martha's last night? Did the
presentation happen or was it postponed for lack of
suitable venue?
___
gnhlug-discuss mailing list
gnhlug-discuss@mail.gnhlug.org
http://mail.gnhlug.org/mailman/listinfo/gnhlug-discuss/
On 6/22/07, Michael ODonnell [EMAIL PROTECTED] wrote:
Worst case, you can open using O_DIRECT, which specifies direct
read/right access. The only limitation really is the reads/writes
need to be aligned to the block size of the device.
Better would be to tell the OS it's OK to cache it,
after *each* write to memory. Nothing that's yet been mentioned
here will yield that sort of behavior. I suppose it would be
possible to simulate it by (say) marking the region of memory
in question as Read Only and then having the SEGFAULT handler
update the corresponding block on the
Tom Buskey [EMAIL PROTECTED] writes:
Digital changed the name too which broke fewer scripts. I liked the way it
mixed BSDisms and SysVisms.
Today they call a system which does this Linux :)
Though, it's amusing to see that *BSD now has /etc/init.d/* as well.
They just spell it differently:
Michael ODonnell
I suppose it would be possible to simulate it by (say) marking the
region of memory in question as Read Only and then having the
SEGFAULT handler update the corresponding block on the device,
but... ick!
You just described ObjectStore:
Christopher Schmidt [EMAIL PROTECTED] writes:
On Thu, Jun 21, 2007 at 10:14:12PM -0400, Paul Lussier wrote:
This was sent to the SAGE members list today. I urge you to watch it.
Offered without comment, except that my politics
are generally those of 1880:
On 6/22/07, Michael ODonnell [EMAIL PROTECTED] wrote:
I was taking about using it with a flash device. If you utilize
O_DIRECT, your write will occur directly to the device, in this
case, written to flash.
(sigh) I have to get back to work, so one last try:
You're confused about
Personally, I found it eye-opening, but was expecting to see it on /.
instead of this list.
There are a number of software employees on this list that might find it
interesting.
I'm not saying it should or should not be on this list. I am thanking
Paul for bringing it to *my* attention; sorry if
On 6/22/07, Paul Lussier [EMAIL PROTECTED] wrote:
This is for a cluster of systems. We don't want to have to depend
upon UPS notification (FedEX maybe ;) and then attempting to shut down
cleanly. We also can't depend upon the customer deploying our
solution with a correct UPS configuration.
24 matches
Mail list logo
