reserve memory for specific usage

Hi
I want to develop a ramdisk, I have tried to allocate memory by
calling ExAllocatePool with NonPagedPool parameter. However in this
way, it will fail if you allocate too much.
So is there a way on a 4 G memory system I reserve 2 Gb memory just
used by my driver?

Are there some expertise ?

Thanks

No. That would be the entire kernel address space.

Even if you only wished to use considerably less, you’d be unlikely to be
able to do this as a single allocation.

You’ll need to do this piecewise using MDL’s.

Good luck,

mm

-----Original Message-----
From: xxxxx@lists.osr.com
[mailto:xxxxx@lists.osr.com] On Behalf Of ??
Sent: Thursday, April 07, 2011 11:00 PM
To: Windows System Software Devs Interest List
Subject: [ntdev] reserve memory for specific usage

Hi
I want to develop a ramdisk, I have tried to allocate memory by calling
ExAllocatePool with NonPagedPool parameter. However in this way, it will
fail if you allocate too much.
So is there a way on a 4 G memory system I reserve 2 Gb memory just used by
my driver?

Are there some expertise ?

Thanks


NTDEV is sponsored by OSR

For our schedule of WDF, WDM, debugging and other seminars visit:
http://www.osr.com/seminars

To unsubscribe, visit the List Server section of OSR Online at
http://www.osronline.com/page.cfm?name=ListServer

Hi,
I knows I need to using MDL’s, I can allocate multi times each time
allocate small size. However in this way, the memory still allocated
from the system.
And the memory can allocated from the system seems are limited depend
on the host memory size.

My problem is whether I can reserve some memory at boot time not
managed by the kernel.

Thanks

On Fri, Apr 8, 2011 at 11:14 AM, Martin O’Brien
wrote:
> No. ?That would be the entire kernel address space.
>
> Even if you only wished to use considerably less, you’d be unlikely to be
> able to do this as a single allocation.
>
> You’ll need to do this piecewise using MDL’s.
>
>
>
> Good luck,
>
> mm
>
>
> -----Original Message-----
> From: xxxxx@lists.osr.com
> [mailto:xxxxx@lists.osr.com] On Behalf Of ??
> Sent: Thursday, April 07, 2011 11:00 PM
> To: Windows System Software Devs Interest List
> Subject: [ntdev] reserve memory for specific usage
>
> Hi
> I want to develop a ramdisk, I have tried to allocate memory by calling
> ExAllocatePool with NonPagedPool parameter. However in this way, it will
> fail if you allocate too much.
> So is there a way on a 4 G memory system I reserve 2 Gb memory just used by
> my driver?
>
> Are there some expertise ??
>
> Thanks
>
> —
> NTDEV is sponsored by OSR
>
> For our schedule of WDF, WDM, debugging and other seminars visit:
> http://www.osr.com/seminars
>
> To unsubscribe, visit the List Server section of OSR Online at
> http://www.osronline.com/page.cfm?name=ListServer
>
>
> —
> NTDEV is sponsored by OSR
>
> For our schedule of WDF, WDM, debugging and other seminars visit:
> http://www.osr.com/seminars
>
> To unsubscribe, visit the List Server section of OSR Online at http://www.osronline.com/page.cfm?name=ListServer
>
>

The list archive contains many discussions of terrible schemes to reserve
physical memory.

Mark Roddy

On Fri, Apr 8, 2011 at 3:45 AM, 钢陈 wrote:

> Hi,
> I knows I need to using MDL’s, I can allocate multi times each time
> allocate small size. However in this way, the memory still allocated
> from the system.
> And the memory can allocated from the system seems are limited depend
> on the host memory size.
>
> My problem is whether I can reserve some memory at boot time not
> managed by the kernel.
>
> Thanks
>
> On Fri, Apr 8, 2011 at 11:14 AM, Martin O’Brien
> wrote:
> > No. That would be the entire kernel address space.
> >
> > Even if you only wished to use considerably less, you’d be unlikely to be
> > able to do this as a single allocation.
> >
> > You’ll need to do this piecewise using MDL’s.
> >
> >
> >
> > Good luck,
> >
> > mm
> >
> >
> > -----Original Message-----
> > From: xxxxx@lists.osr.com
> > [mailto:xxxxx@lists.osr.com] On Behalf Of ??
> > Sent: Thursday, April 07, 2011 11:00 PM
> > To: Windows System Software Devs Interest List
> > Subject: [ntdev] reserve memory for specific usage
> >
> > Hi
> > I want to develop a ramdisk, I have tried to allocate memory by calling
> > ExAllocatePool with NonPagedPool parameter. However in this way, it will
> > fail if you allocate too much.
> > So is there a way on a 4 G memory system I reserve 2 Gb memory just used
> by
> > my driver?
> >
> > Are there some expertise ?
> >
> > Thanks
> >
> > —
> > NTDEV is sponsored by OSR
> >
> > For our schedule of WDF, WDM, debugging and other seminars visit:
> > http://www.osr.com/seminars
> >
> > To unsubscribe, visit the List Server section of OSR Online at
> > http://www.osronline.com/page.cfm?name=ListServer
> >
> >
> > —
> > NTDEV is sponsored by OSR
> >
> > For our schedule of WDF, WDM, debugging and other seminars visit:
> > http://www.osr.com/seminars
> >
> > To unsubscribe, visit the List Server section of OSR Online at
> http://www.osronline.com/page.cfm?name=ListServer
> >
> >
>
> —
> NTDEV is sponsored by OSR
>
> For our schedule of WDF, WDM, debugging and other seminars visit:
> http://www.osr.com/seminars
>
> To unsubscribe, visit the List Server section of OSR Online at
> http://www.osronline.com/page.cfm?name=ListServer
>

钢陈 wrote:

Hi,
I knows I need to using MDL’s, I can allocate multi times each time
allocate small size. However in this way, the memory still allocated
from the system.
And the memory can allocated from the system seems are limited depend
on the host memory size.

My problem is whether I can reserve some memory at boot time not
managed by the kernel.

Why do you want a RAM disk? The operating system already uses up to
half of your physical memory for disk buffers, and it is much smarter
about that then you are. If you are hoping to improve performance, you
are wrong.

Really, this operating system has had 20 years of research and
development in performance optimization. There is very little you could
come up with in less than a year to improve things measurably.


Tim Roberts, xxxxx@probo.com
Providenza & Boekelheide, Inc.

Gang wrote …

I knows I need to using MDL’s, I can allocate multi times each time allocate small size. However in this way, the memory still allocated from the system. And the memory can allocated from the system seems are limited depend on the host memory size. My problem is whether I can reserve some memory at boot time not managed by the kernel.

… and I replied …

Umm, pretty much *all* memory at boot time is managed by the kernel – if it’s addressable then HAL has control of it, if it’s not addressable (say memory beyond 4GB on a 32bit machine with 8GB in DRAM) then that’s not addressable period. If it’s memory that’s on a PCI device then that memory is exposed by the hardware and through the OS, again under the control of the OS.

What specific segment of memory are you attempting to reserve from, and what is your basis of belief that that segment is not managed by the kernel?

Cheers!

choward@ix.netcom.com wrote:

Umm, pretty much *all* memory at boot time is managed by the kernel – if it’s addressable then HAL has control of it, if it’s not addressable (say memory beyond 4GB on a 32bit machine with 8GB in DRAM) then that’s not addressable period. If it’s memory that’s on a PCI device then that memory is exposed by the hardware and through the OS, again under the control of the OS.

What specific segment of memory are you attempting to reserve from, and what is your basis of belief that that segment is not managed by the kernel?

Well, it is certainly possible to use the /BURNMEM switch in boot.ini
(or it’s BCD equivalent) to tell the system to ignore a region of
memory. A driver can then go out and use that memory for its own purposes.

The problem with that is that it’s the wild, wild west. There is no
central control over that memory, so there’s no way to ensure that three
drivers don’t all use it for themselves, resulting in disaster. That’s
exactly what Don was warning about. We’ve had this discussion before.
It can be done, but it cannot be done safely.


Tim Roberts, xxxxx@probo.com
Providenza & Boekelheide, Inc.

Tim Roberts skrev 2011-04-08 19:39:

Why do you want a RAM disk? The operating system already uses up to
half of your physical memory for disk buffers, and it is much smarter
about that then you are. If you are hoping to improve performance, you
are wrong.

How clever is the OS actually about disk buffers? I have been thinking
about trying to use a RAM disk to improve build performance.

We run build processes on a specific server. The process starts by
checking out about 100 megs of source codes from the repository. Then
the build runs, creating somewhere around 1-2 gigabytes of temporary
files in the process. Then a cleanup script runs and deletes everything
but the executables, symbol files and the installer, leaving currently
106 megs of data on disk.

It seems to me that the OS is actually writing the data for the
temporary files to disk, even after they have been deleted. That is why
I have been considering setting up a ram disk for the build process to
use for all the temporary files. Does that make sense or am I just
misinterpreting the disk io? I am not sure whether the file system will
cancel writing out file data for deleted files to disk?

Andreas Hansson

Andreas Hansson wrote:

How clever is the OS actually about disk buffers? I have been thinking
about trying to use a RAM disk to improve build performance.

We run build processes on a specific server. The process starts by
checking out about 100 megs of source codes from the repository. Then
the build runs, creating somewhere around 1-2 gigabytes of temporary
files in the process. Then a cleanup script runs and deletes everything
but the executables, symbol files and the installer, leaving currently
106 megs of data on disk.

It seems to me that the OS is actually writing the data for the
temporary files to disk, even after they have been deleted.

OK, so I’m being overly grumpy. Buffering helps enormously with reading
– not so much with writing.

Your situation is crying out for one of the new solid state disks.
Their performance has to be seen to be believed. With an SSD as the
system drive, Windows 7 is logged in about 15 seconds after the BIOS is
done. That way, you aren’t taking away any of your precious system RAM,
which is critical for the build process.

That is why
I have been considering setting up a ram disk for the build process to
use for all the temporary files. Does that make sense or am I just
misinterpreting the disk io? I am not sure whether the file system will
cancel writing out file data for deleted files to disk?

It will not. Pending write data is committed when the file is closed.


Tim Roberts, xxxxx@probo.com
Providenza & Boekelheide, Inc.

“Tim Roberts” <> wrote in message news:xxxxx@ntdev…

Your situation is crying out for one of the new solid state disks.
Their performance has to be seen to be believed.

Then what the OP wants can be described as a “cheap virtual SSD” ?

“Andreas Hansson” <> wrote in message news:…

>> Does that make sense or am I just
>> misinterpreting the disk io? I am not sure whether the file system will
>> cancel writing out file data for deleted files to disk?

Have you tested your workload with a real SSD to see if it helps
significantly at all?

– pa

Stick a good quality flash drive in a USB slot and use that to accelerate your performance… Or put the temp build files on it.

Cheap way to get a (small) ssd.

Peter
OSR

Tim, as far as I know, this isn’t quite always the case (at least on win7). Then again, I’m not a filesystems nor NTFS expert so perhaps someone will correct me (but empirical evidence seems to disagree with that assertion).

For NTFS, at least on win7 and based on my understanding, we – today, which might or might not be subject to change based in the future – flush the cache (synchronously) on handle close if the file should happen to reside on removable media. If it doesn’t reside on removable media, barring various other circumstances, one can delete the file later and we’ll then discard the modified pages (i.e. cease writing them out to disk), should the last reference to the file go away.

(One could also set the file delete disposition via SetFileInformationByHandle before closing the file and skip writing the modified pages out, or truncate the file before closing the file, similarly.)

The simple test for this is to create a large file on a non-removable NTFS volume (say, dump 1GB into it), then close the handle, and watch the modified page count go down with disk I/O as the file is written out (easy to do in the Memory tab of Resource Monitor on win7). You can delete the file manually and watch the modified pages be immediately removed (with the I/O associated with writing them out ceasing).

Andreas ( Hello! :slight_smile: Fancy spotting you here… ) –

What OS releases have you tried this on (have you attempted win7/srv08r2 in particular)? Are there any extraneous filesystem filters in the mix that might be altering the behavior here (AV software, etc)?

Note that memory pressure could always induce heavy writing out of modified pages if available physical memory runs low (of course, this would mean bad news for your RAM disk too).

  • S (Msft)

-----Original Message-----
From: xxxxx@lists.osr.com [mailto:xxxxx@lists.osr.com] On Behalf Of Tim Roberts
Sent: Friday, April 08, 2011 11:48 AM
To: Windows System Software Devs Interest List
Subject: Re: [ntdev] reserve memory for specific usage

Andreas Hansson wrote:

How clever is the OS actually about disk buffers? I have been thinking
about trying to use a RAM disk to improve build performance.

We run build processes on a specific server. The process starts by
checking out about 100 megs of source codes from the repository. Then
the build runs, creating somewhere around 1-2 gigabytes of temporary
files in the process. Then a cleanup script runs and deletes
everything but the executables, symbol files and the installer,
leaving currently
106 megs of data on disk.

It seems to me that the OS is actually writing the data for the
temporary files to disk, even after they have been deleted.

OK, so I’m being overly grumpy. Buffering helps enormously with reading
– not so much with writing.

Your situation is crying out for one of the new solid state disks.
Their performance has to be seen to be believed. With an SSD as the system drive, Windows 7 is logged in about 15 seconds after the BIOS is done. That way, you aren’t taking away any of your precious system RAM, which is critical for the build process.

That is why
I have been considering setting up a ram disk for the build process to
use for all the temporary files. Does that make sense or am I just
misinterpreting the disk io? I am not sure whether the file system
will cancel writing out file data for deleted files to disk?

It will not. Pending write data is committed when the file is closed.


Tim Roberts, xxxxx@probo.com
Providenza & Boekelheide, Inc.


NTDEV is sponsored by OSR

For our schedule of WDF, WDM, debugging and other seminars visit:
http://www.osr.com/seminars

To unsubscribe, visit the List Server section of OSR Online at http://www.osronline.com/page.cfm?name=ListServer

Hmm… maybe, maybe not. Remember that USB2 storage devices are likely to top out at around 20MB/s for bulk transfer performance in ideal cases.

If one is really completely bound by seeking, there may be a benefit to be gleaned here versus spinning platter, but otherwise I suspect that there’s a real chance that performance may well be notably worse (versus a dedicated SATA, high speed disk), even factoring in reduced load on the spinning platter media. If nothing else, there’s going to be at least nearly two minutes burned to write 2GB of temporary files out to a USB2 device given an ideal bulk transfer scenario… (were one to assume that they would be all written).

  • S

-----Original Message-----
From: xxxxx@lists.osr.com [mailto:xxxxx@lists.osr.com] On Behalf Of xxxxx@osr.com
Sent: Friday, April 08, 2011 7:03 PM
To: Windows System Software Devs Interest List
Subject: RE:[ntdev] reserve memory for specific usage

Stick a good quality flash drive in a USB slot and use that to accelerate your performance… Or put the temp build files on it.

Cheap way to get a (small) ssd.

Peter
OSR


NTDEV is sponsored by OSR

For our schedule of WDF, WDM, debugging and other seminars visit:
http://www.osr.com/seminars

To unsubscribe, visit the List Server section of OSR Online at http://www.osronline.com/page.cfm?name=ListServer

I would suggest using xperf to understand where exactly these I/Os are coming from.

xperf -on base
wait 1 minute while IO is happening
xperf -d build_disk_io.etl
xperf build_disk_io.etl
right-click the disk usage graph, choose “Summary table”.

Thanks,
Pavel

-----Original Message-----
From: xxxxx@lists.osr.com [mailto:xxxxx@lists.osr.com] On Behalf Of Skywing
Sent: Friday, April 08, 2011 7:22 PM
To: Windows System Software Devs Interest List
Subject: RE: [ntdev] reserve memory for specific usage

Tim, as far as I know, this isn’t quite always the case (at least on win7). Then again, I’m not a filesystems nor NTFS expert so perhaps someone will correct me (but empirical evidence seems to disagree with that assertion).

For NTFS, at least on win7 and based on my understanding, we – today, which might or might not be subject to change based in the future – flush the cache (synchronously) on handle close if the file should happen to reside on removable media. If it doesn’t reside on removable media, barring various other circumstances, one can delete the file later and we’ll then discard the modified pages (i.e. cease writing them out to disk), should the last reference to the file go away.

(One could also set the file delete disposition via SetFileInformationByHandle before closing the file and skip writing the modified pages out, or truncate the file before closing the file, similarly.)

The simple test for this is to create a large file on a non-removable NTFS volume (say, dump 1GB into it), then close the handle, and watch the modified page count go down with disk I/O as the file is written out (easy to do in the Memory tab of Resource Monitor on win7). You can delete the file manually and watch the modified pages be immediately removed (with the I/O associated with writing them out ceasing).

Andreas ( Hello! :slight_smile: Fancy spotting you here… ) –

What OS releases have you tried this on (have you attempted win7/srv08r2 in particular)? Are there any extraneous filesystem filters in the mix that might be altering the behavior here (AV software, etc)?

Note that memory pressure could always induce heavy writing out of modified pages if available physical memory runs low (of course, this would mean bad news for your RAM disk too).

  • S (Msft)

-----Original Message-----
From: xxxxx@lists.osr.com [mailto:xxxxx@lists.osr.com] On Behalf Of Tim Roberts
Sent: Friday, April 08, 2011 11:48 AM
To: Windows System Software Devs Interest List
Subject: Re: [ntdev] reserve memory for specific usage

Andreas Hansson wrote:

How clever is the OS actually about disk buffers? I have been thinking
about trying to use a RAM disk to improve build performance.

We run build processes on a specific server. The process starts by
checking out about 100 megs of source codes from the repository. Then
the build runs, creating somewhere around 1-2 gigabytes of temporary
files in the process. Then a cleanup script runs and deletes
everything but the executables, symbol files and the installer,
leaving currently
106 megs of data on disk.

It seems to me that the OS is actually writing the data for the
temporary files to disk, even after they have been deleted.

OK, so I’m being overly grumpy. Buffering helps enormously with reading
– not so much with writing.

Your situation is crying out for one of the new solid state disks.
Their performance has to be seen to be believed. With an SSD as the system drive, Windows 7 is logged in about 15 seconds after the BIOS is done. That way, you aren’t taking away any of your precious system RAM, which is critical for the build process.

That is why
I have been considering setting up a ram disk for the build process to
use for all the temporary files. Does that make sense or am I just
misinterpreting the disk io? I am not sure whether the file system
will cancel writing out file data for deleted files to disk?

It will not. Pending write data is committed when the file is closed.


Tim Roberts, xxxxx@probo.com
Providenza & Boekelheide, Inc.


NTDEV is sponsored by OSR

For our schedule of WDF, WDM, debugging and other seminars visit:
http://www.osr.com/seminars

To unsubscribe, visit the List Server section of OSR Online at http://www.osronline.com/page.cfm?name=ListServer


NTDEV is sponsored by OSR

For our schedule of WDF, WDM, debugging and other seminars visit:
http://www.osr.com/seminars

To unsubscribe, visit the List Server section of OSR Online at http://www.osronline.com/page.cfm?name=ListServer

My purpose to buildup a ramdisk is the same with you.
And I also want to know is there a way windows can support reserve
memory for specific usage

On Sat, Apr 9, 2011 at 2:25 AM, Andreas Hansson
wrote:
> Tim Roberts skrev 2011-04-08 19:39:
>>
>> Why do you want a RAM disk? ?The operating system already uses up to
>> half of your physical memory for disk buffers, and it is much smarter
>> about that then you are. ?If you are hoping to improve performance, you
>> are wrong.
>
> How clever is the OS actually about disk buffers? I have been thinking about
> trying to use a RAM disk to improve build performance.
>
> We run build processes on a specific server. The process starts by checking
> out about 100 megs of source codes from the repository. Then the build runs,
> creating somewhere around 1-2 gigabytes of temporary files in the process.
> Then a cleanup script runs and deletes everything but the executables,
> symbol files and the installer, leaving currently 106 megs of data on disk.
>
> It seems to me that the OS is actually writing the data for the temporary
> files to disk, even after they have been deleted. That is why I have been
> considering setting up a ram disk for the build process to use for all the
> temporary files. Does that make sense or am I just misinterpreting the disk
> io? I am not sure whether the file system will cancel writing out file data
> for deleted files to disk?
>
>
> Andreas Hansson
>
> —
> NTDEV is sponsored by OSR
>
> For our schedule of WDF, WDM, debugging and other seminars visit:
> http://www.osr.com/seminars
>
> To unsubscribe, visit the List Server section of OSR Online at
> http://www.osronline.com/page.cfm?name=ListServer
>
>

On Saturday 09 Apr 2011 03:21:30 Skywing wrote:

For NTFS, at least on win7 and based on my understanding, we – today,
which might or might not be subject to change based in the future – flush
the cache (synchronously) on handle close if the file should happen to
reside on removable media. If it doesn’t reside on removable media,
barring various other circumstances, one can delete the file later and
we’ll then discard the modified pages (i.e. cease writing them out to
disk), should the last reference to the file go away.

The documentation for CreateFile confirms that.
From http://msdn.microsoft.com/en-
us/library/aa363858%28v=vs.85%29.aspx#caching_behavior:

“Specifying the FILE_ATTRIBUTE_TEMPORARY attribute causes file systems to avoid
writing data back to mass storage if sufficient cache memory is available,
because an application deletes a temporary file after a handle is closed. In
that case, the system can entirely avoid writing the data. Although it does
not directly control data caching in the same way as the previously mentioned
flags, the FILE_ATTRIBUTE_TEMPORARY attribute does tell the system to hold as
much as possible in the system cache without writing and therefore may be of
concern for certain applications.”


Bruce Cran

The readyboost test is a clue if your USB stick is worth bothering with.

On Friday, April 8, 2011, Skywing wrote:
> Hmm… maybe, maybe not. ?Remember that USB2 storage devices are likely to top out at around 20MB/s for bulk transfer performance in ideal cases.
>
> If one is really completely bound by seeking, there may be a benefit to be gleaned here versus spinning platter, but otherwise I suspect that there’s a real chance that performance may well be notably worse (versus a dedicated SATA, high speed disk), even factoring in reduced load on the spinning platter media. ?If nothing else, there’s going to be at least nearly two minutes burned to write 2GB of temporary files out to a USB2 device given an ideal bulk transfer scenario… (were one to assume that they would be all written).
>
> - S
>
>
> -----Original Message-----
> From: xxxxx@lists.osr.com [mailto:xxxxx@lists.osr.com] On Behalf Of xxxxx@osr.com
> Sent: Friday, April 08, 2011 7:03 PM
> To: Windows System Software Devs Interest List
> Subject: RE:[ntdev] reserve memory for specific usage
>
> Stick a good quality flash drive in a USB slot and use that to accelerate your performance… Or put the temp build files on it.
>
> Cheap way to get a (small) ssd.
>
> Peter
> OSR
>
>
> —
> NTDEV is sponsored by OSR
>
> For our schedule of WDF, WDM, debugging and other seminars visit:
> http://www.osr.com/seminars
>
> To unsubscribe, visit the List Server section of OSR Online at http://www.osronline.com/page.cfm?name=ListServer
>
> —
> NTDEV is sponsored by OSR
>
> For our schedule of WDF, WDM, debugging and other seminars visit:
> http://www.osr.com/seminars
>
> To unsubscribe, visit the List Server section of OSR Online at http://www.osronline.com/page.cfm?name=ListServer
>


Mark Roddy

“Mark Roddy” wrote in message news:xxxxx@ntdev…
> The readyboost test is a clue if your USB stick is worth bothering with.

But USB disk is removable, therefore writeback will occur when files on it
are closed?
The goal is to keep them in the cache until they are deleted (which, as the
OP says, will happen soon).

Is there some way (a policy? registry tweak?) to force
FILE_ATTRIBUTE_TEMPORARY on all
files written under specified path?
If not, can this be done with a simple FS filter?

Regards,
–pa

> On Friday, April 8, 2011, Skywing wrote:
>> Hmm… maybe, maybe not. Remember that USB2 storage devices are likely
>> to top out at around 20MB/s for bulk transfer performance in ideal cases.
>>
>> If one is really completely bound by seeking, there may be a benefit to
>> be gleaned here versus spinning platter, but otherwise I suspect that
>> there’s a real chance that performance may well be notably worse (versus
>> a dedicated SATA, high speed disk), even factoring in reduced load on the
>> spinning platter media. If nothing else, there’s going to be at least
>> nearly two minutes burned to write 2GB of temporary files out to a USB2
>> device given an ideal bulk transfer scenario… (were one to assume that
>> they would be all written).
>>
>> - S
>>
>>
>> -----Original Message-----
>> From: xxxxx@lists.osr.com
>> [mailto:xxxxx@lists.osr.com] On Behalf Of xxxxx@osr.com
>> Sent: Friday, April 08, 2011 7:03 PM
>> To: Windows System Software Devs Interest List
>> Subject: RE:[ntdev] reserve memory for specific usage
>>
>> Stick a good quality flash drive in a USB slot and use that to accelerate
>> your performance… Or put the temp build files on it.
>>
>> Cheap way to get a (small) ssd.
>>
>> Peter
>> OSR

You can change the props in device manager to treat the disk as fixed (at your own risk for data loss if you just yank it without going through the safe remove hw ui). Don’t know if the settings it stores are documented or not.

d

debt from my phone

-----Original Message-----
From: Pavel A.
Sent: Saturday, April 09, 2011 3:37 PM
To: Windows System Software Devs Interest List
Subject: Re:[ntdev] reserve memory for specific usage

“Mark Roddy” wrote in message news:xxxxx@ntdev…
> The readyboost test is a clue if your USB stick is worth bothering with.

But USB disk is removable, therefore writeback will occur when files on it
are closed?
The goal is to keep them in the cache until they are deleted (which, as the
OP says, will happen soon).

Is there some way (a policy? registry tweak?) to force
FILE_ATTRIBUTE_TEMPORARY on all
files written under specified path?
If not, can this be done with a simple FS filter?

Regards,
–pa

> On Friday, April 8, 2011, Skywing wrote:
>> Hmm… maybe, maybe not. Remember that USB2 storage devices are likely
>> to top out at around 20MB/s for bulk transfer performance in ideal cases.
>>
>> If one is really completely bound by seeking, there may be a benefit to
>> be gleaned here versus spinning platter, but otherwise I suspect that
>> there’s a real chance that performance may well be notably worse (versus
>> a dedicated SATA, high speed disk), even factoring in reduced load on the
>> spinning platter media. If nothing else, there’s going to be at least
>> nearly two minutes burned to write 2GB of temporary files out to a USB2
>> device given an ideal bulk transfer scenario… (were one to assume that
>> they would be all written).
>>
>> - S
>>
>>
>> -----Original Message-----
>> From: xxxxx@lists.osr.com
>> [mailto:xxxxx@lists.osr.com] On Behalf Of xxxxx@osr.com
>> Sent: Friday, April 08, 2011 7:03 PM
>> To: Windows System Software Devs Interest List
>> Subject: RE:[ntdev] reserve memory for specific usage
>>
>> Stick a good quality flash drive in a USB slot and use that to accelerate
>> your performance… Or put the temp build files on it.
>>
>> Cheap way to get a (small) ssd.
>>
>> Peter
>> OSR


NTDEV is sponsored by OSR

For our schedule of WDF, WDM, debugging and other seminars visit:
http://www.osr.com/seminars

To unsubscribe, visit the List Server section of OSR Online at http://www.osronline.com/page.cfm?name=ListServer

> Stick a good quality flash drive in a USB slot and use that to accelerate your performance…

25MB/s on USB2 vs 150MB/s on SATA? a win?


Maxim S. Shatskih
Windows DDK MVP
xxxxx@storagecraft.com
http://www.storagecraft.com