Archived from groups: microsoft.public.win2000.file_system (
More info?)
In article <#ly775qNEHA.624@TK2MSFTNGP11.phx.gbl>,
andrea catto' <acatto@dataflight.com> wrote:
>here's what I learned my self empirically,
>
>'IT DEPENDS'
>
>compression is NOT always working, in which case it's a waste of CPU versus,
>no gain, slower for nothing bascially.
>
>some times it does work but minimally, not a lot of gain.
>
>in some cases it work very well, for instance in text files, or easily
>compressible data, in which case here's what I have seen.
> in case of frequent access, it may deteriorate performances because of
>the CPU time taken to compress/decompress
> and in other cases I found compression 'could' actually improve disk
>usage immensly and YET improve timing too.
>
>let me describe this 'paradox' of total improvement, if a file can be
>compressed, the clusters involved in read-write are also less then normally
>needed, therefore in this case it coult be beneficial to use CPU time versus
>disk time, because CPU is faster in most cases then disk.
I differ.
I started using NTFS compression on entire C drives when our PCs were
486/66 systems and we were doing very CPU-intensive stuff. If there
was any CPU cost we would have noticed in, believe me. I spent lots of
time looking at perfmon data.
Today's CPUs are literally 100s of times faster, and generally
underused. There's plenty of unused cycles foating around. Most IDE
controllers use lots of CPU cycles to move data to/from the disk. If
you can reduce the # of sectors to read you reduce the CPU time to
read that file by a similar amount. This wins big-time for highly
compressed files.
If you are storing huge files that are already compressed (mp3, mpeg)
you won't get any space from compression, but I'm not sure that the
cost of the having compression tuned on is measurable. These days
I work with huge TIFFs (60MB) in Photoshop. My TIFFs don't compress
much and I don't see any win, or loss, to file system compression. I
haven't played with PSD files yet, which may be even bigger.
>
>finally.
>you might want to compress data that is not commonly/frequently used, such
>as a whole directory branch.
>or could also want to 'try' compressing certain files or directories that
>contain text or stuff you know is quite reduntant in its content and
>possibly over the size of many clusters to see some benefits.
>
>
>
>"garrydt" <anonymous@discussions.microsoft.com> wrote in message
>news:a0e901c4348d$eaf9f7a0$a601280a@phx.gbl...
>> Does enabling compression on your C:/ drive impact
>> performance; if so is it a real killer?
>
>
--
Al Dykes
-----------
adykes at p a n i x . c o m