Gerry.
Gerry Cornell said:
Mak
Files are not fragmented by being accessed. They become fragmented when
they are changed.
"Accessed" does not automatically mean "accessed with read function", Gerry.
"Accessed" was used as opposite to 'not accessed', 'left alone'. To change
the file, you need to access it first. Does that make sense now?
Rewriting files leaves available free space scattered all over
the drive, where there is a no partitioning .
(Semantics: we are talking about multiple partitions vs. single partition
(still a partition)). I have no problem with free space all over the drive.
What I do have problem with is severely fragmented free space. More room to
breath on single partition makes more room for contiguous free space.
This means that any new large file
can be significantly fragmented when it is first written.
There is no problem with the number of fragments, Gerry. The problem is only
with the number of split I/Os compared to total I/O made to "access" the
file. As I was trying to explain to you: sequential access is very, very
rare. So, we have a severely fragmented file. When the file is accessed it's
normally _not_ read from the beginning to the end in one large I/O
(Something the makers of defrag programs don't tell you, and their synthetic
tests don't show). It's read in chunks, between the I/Os to the above file
Windows will read / write to other files, the heads will move no matter
what. So, it doesn't matter if file is contiguous or not, it matters
however, how close (something that Prefetch does, or smart defrag with file
placement optimization) to the other last file Windows was accessing the
next chunk of the above file is . and how many times I/Os were split because
I/O has to cross the border of 2 fragments. If that split I/O number
_compared_ (as a percentage for example) to total number of I/Os is
insignificant, so will be the performance benefit of running defrag.
You need less free space
on a partitition containing windows and programmes or archived data, where
changes occur occasionally
<sarcasm on> Nothing is written to %windir% and %programfiles% during
operation of a computer, ever. <sarcasm off>
You know how often stuff got written to %windir% / %programfiles% for a
fact?
and more free space on a current data
but not as much as there can be with single partition.
or a virtual
memory partition
WTF is "virtual memory partition"? Windows doesn't have that.
But if you are talking (big stretch, paging file is not virtual memory)
about a partition where all by itself, your paging file lives - that is a
big no-no, bad, stupid thing to do.
You want your paging file in the middle of your most used partition of your
least used physical drive. Since we are talking about single drive here, you
want your paging file to be in the middle of your OS + Program Files
partition (if you're still making multiple partitions).
where changes are taking place all the time.
Actually it can be the reverse because one should allocate a higher
percentage
free space to a data partition thanto a windows / programme partition.
One should do no such thing. One should stop and think about reasons for
multiple partitions first, before looking up the price of PM.
I have used FileMon. I would not dream of trying to draw such a conclusion
from
a programme that deluges the user with so much information. You cannot
see the
wood for the trees.
There are no problems with info in FileMon, besides, there are built-in
filters when "so much information" is present.
Hard drive will usually become obsolete before they wear out. That's why
many
manufacturers offer a 3 year warranty
You've got that backwards. 3 years are offered because of expected lifespan,
not the other way around. I have few 2GB SCSI drives (proxy cache) that are
close to celebrating their 7th birthday. I will not replace them til they
die.
I am with you there.
That is what I am doing. The fragments are closer together in a dedicated
partition than
they would be in a single partition when they would be scattered all the
disk !
See above on the reading patterns, that's not what you are doing /
promoting.
Exactly! Close to each other.Thank you for arguing my case so well <g>.
Read: close to other frequently accessed files as well (reading patterns) -
your setup lacks that.
The size of files is a big factor. Large files when written will naturally
break
into many fragments because of the size of available free space. The
System
Restore folders are a good illustration of what I mean.
They will not break "naturally", but for the reason: lack of contiguous free
space. Windows has no evil intentions fragmenting your big files.
I can only reply on the basis of what you say not what I think you
intended to say.
If you want to continue to advance that as an argument you will have to
restate
you case omitting everything than relies on the RAID technology.
I don't "have to", Your Honour, you, however, _may_ make some efforts where
/ if they are due to understand the example.
Why do people bombard others with spam and nasties. Some will make up
numbers but more frequently persons will put forward statements including
numbers favourable to their objective and omitting those which are
unhelpful.
You can measure individual performances, although the results are applied
in
the wrong way they can be demotivating and you end up with wrong outcome.
Relevance?
What I'm saying / suggesting is this: do the tests, keep the numbers and try
not lie to yourself. (You _don't_have_to_ do the tests, I'm not seeking to
convert you at all - it's not important.)
I made the point to demonstrate that often all is not what it seems to be.
There wasn't a need for one, considering I'm not particularly interested in
your numbers, I though you will be, it looks like I was wrong.
Partition Magic has never let me down.
Never is such a long time, Gerry.
You are not the only one who uses PM.
I can say: PM will never let me down, can you?
Not looking for one ATM. Simply having a discussion. <- scratch that if it
just a signature.>