Re: Re:What's the longest time before Windows flushes data to disk?

Tech-Archive recommends: Fix windows errors by optimizing your registry

Don, thank you for the direct answer by the way. What you're saying is that
essentially there is no upper bound on how long before the cache gets
flushed, which is exactly the type of info I was looking for. The point
about designing the application to flush or not buffer the data is well
taken and it's a given that our app will need to be changed. Still, if you
could help me visualise more precisely how the cache operates, that would be
great, if you have the time, as I could be more confident about the chances
that this really was the problem in this case. For example:

Does the Windows cache work strictly in a FIFO fashion (I know the internal
hard disk cache doesn't, but how about the Windows cache)?

If some application continuously reads the hard disk, how is writing to disk
prioritised when another application that is writing is not asking for any
special treatment, e.g. it is not making it's files "write through" nor
flushing the cache. As long as there is enough memory, does reading from
disk always take precedence?

20 days must surely be an extreme case, right? Also, what do you mean? That
a particular data item has stayed in cache and not been flushed for 20 days?
That all data was in cache for 20 days (e.g. because the hard disk failed)?
That the cache as a whole was never fully flushed for 20 days? In the last
case, what was the longest time that any one particular item stayed in
cache, without being written to disk, something close to 20 days or
considerably shorter?

What I'd love to see, if such a thing exists, is a statistic that plots the
percentage of data items against their age in the write cache at various
points in time during the boot process or other "busy" time of a "typical"
Windows 2003 server.

I know I'm asking a lot. However any answers, be they statistics (which
would be wonderful), technical insight into the cache or anecdotes will be
appreciated.

On the other hand (Anton), please do not ask me for more details about the
specific case. This is mainly because I wasn't there and I don't have a lot
of details. You cannot help troubleshoot or judge the specific case and I'm
not asking you to. I just wanted to give a little bit of background to put
this into context.


"MB" <bla@xxxxxxx> wrote in message
news:u7FZh87YGHA.4884@xxxxxxxxxxxxxxxxxxxxxxx
No kidding? Therefore I shouldn't rely on data being flushed, fair enough,
but ... 20 DAYS? Tell me you're not joking. This has actually happened? I
feel there may be an interesting story here ... care to share?

"Don Burn" <burn@xxxxxxxxxxxxxxxx> wrote in message
news:OqPVWw7YGHA.3424@xxxxxxxxxxxxxxxxxxxxxxx
Depending on the environment and what is happening, I know of systems
that
have not flushed the data in 20 DAYS!. You should not assume that data
is
flushed, unless you actively design the application to flush or not
buffer
in the first place.


--
Don Burn (MVP, Windows DDK)
Windows 2k/XP/2k3 Filesystem and Driver Consulting
Remove StopSpam from the email to reply


.

Relevant Pages

  • Re: Hard drive corrupt at every reboot
    ... when I have upgraded the user to Windows 2000 ... the disk-controller driver used under both Windows 2000 ... hardware cache on the Hard Disk. ...
    (microsoft.public.windowsxp.hardware)
  • Re: Need new computer
    ... Good processor, you'll be able to use 3GB of the 4GB of RAM, giving ... Good sized primary hard disk drive. ... I recommend one with at least a 4MB cache. ... there are advantages to a quad I'd go strongly in that direction. ...
    (alt.sys.pc-clone.dell)
  • Re: Event ID 32 - write cache enabled
    ... The local domain controller could not disable the software-based disk write ... cache on the following hard disk. ... So, it does appear the Windows is attemtping the disable the write cache, ...
    (microsoft.public.windows.server.general)
  • Re: System Spec (OT?)
    ... but the Core2 processors with only 1 or 2MB of cache ... *Memory - 1^st Hard Disk* ... *Graphics Card* ... Sound BlasterAudigySE with 7.1 Surround Sound: ...
    (Ubuntu)
  • Re: New FAT32 performance
    ... Copy from local hard disk to USB hard disk (via File Commander): ... CACHE has space for 4096 sectors ...
    (comp.os.os2.setup.storage)