EWF, immediate commit

[dalai lamah]

Is there a write filter driver similar to EWF, able to repeatedly and
immediately commit the changes when requested? The -live option in EWF
commits immediately, but it's very slow and also disables EWF until the
next reboot, and this isn't good for me.

Actually I would need more a very lazy disk cache rather than a filter
driver, something that commits the disk writes every ten seconds or so; can
I tweak the XPE cache manager changing some registry keys to make it
lazier?

 

[Sean Liming \(eMVP\)]

Have you looked at the new FBWF that is part of FP 2007?:

http://www.seanliming.com/Docs/FBWF.pdf

Regards,

Sean Liming
www.sjjmicro.com / www.seanliming.com
XP Embedded Book Author - XP Embedded Advanced, XP Embedded
SupplementalToolkit

 

[KM]

Sounds like you may need to look at the new component of FP2007 - FBWF:
http://msdn2.microsoft.com/en-us/library/aa940926.aspx

I don't think disk cache can be used instead of a write filter. But you can
read more info about 2K/XP caching here:
http://doc.sch130.nsc.ru/www.sysinternals.com/ntw2k/source/cacheman.shtml

Basically you can control the cache size with the tool from sysinternals.com
(MS): CacheSet
(http://www.microsoft.com/technet/sysinternals/FileAndDisk/CacheSet.mspx).

KM

 

[dalai lamah]

Un bel giorno KM digitò:

Sounds like you may need to look at the new component of FP2007 - FBWF:
http://msdn2.microsoft.com/en-us/library/aa940926.aspx

I wasn't aware of this feature pack, it looks interesting! It isn't very
clear what kind of license it has; is it free (for registered Embedded
Studio users) or is it just an evaluation?

I don't think disk cache can be used instead of a write filter.

Actually there isn't so much difference between a write filter and a modern
OS cache; sure, you have far less control with the cache, for example you
can't prevent every possible write. But if you just need to reduce the
number of writes to increase the CF lifespan, it's just fine.

But you can
read more info about 2K/XP caching here:
http://doc.sch130.nsc.ru/www.sysinternals.com/ntw2k/source/cacheman.shtml

A program like that would be fine; the interesting parameters are
CcIdleDelay and CcDirtyPageThreshold, which control the threshold (in time
and size) after that the lazy writer kicks in.

Unfortunately this kind of utilities use very undocumented kernel
structures, for example Russinovich's CacheMan runs just under NT. Some
interesting informations about CacheMan and lazy writes:

http://www.microsoft.com/technet/prodtechnol/windows2000serv/maintain/optimize/wperfch7.mspx

I think that the dirty page threshold could be set through the registry
(it's in
HKLM\System\CurrentControlSet\Control\SessionManager\MemoryManagement). I
don't know how to set the idle delay though.

 

[KM]

Doh.. The link you posted was the one I wanted to post but accidentally copied the CacheMan link
CacheMan was developed for NT4 and won't run on 2K/XP.

FP 2007 is a free update for current XP Embedded users.

You are probably trying to compare RAM mode of the EWF/FBWF to the disk cache "engine".
Write Filter is more common component that allows you to redirect disk writes to either RAM or disk overlay.
Also, with he WF you are in control when the overlay gets flushed to the disk and hence you are responsible for making the commit
operation graceful.

With the disk cache, regardless of what flush timeout you set up, the OS is doing the data flushing in background and this may
potentially lead you to a damage in FS structure or some data when an accidental power off happened in the middle of such flush.
Also, I was under impression the lazy disk cache writer in XP works not by timeout period but rather always runs in a low (idle)
priority thread flushing whatever got in its buffer that can be set small or large. I might be wrong here since I didn't read enough
about the lazy writer algorithms implemented in XP.