Defrag - Page file - NTFS questions

[yandr]

I was wondering about defrag and the page files.

Defrag

Does anyone have any statistics on the performance gains? Personally, I
think that today's hard disk drives are so fast, that a difference between a
moderately fragmented drive and a defragmented one will be negligible. Has
anybody found reliable information on the net concerning this matter?

When I say moderately, I am refering to fragmentation being the result of an
average usage, ratehr than artificially fragmenting the drive just to prove
that it does have an effect on performance. Now that most systems use (or
ought to be using) NTFS, fragmentation is even less, so the performance
difference should be less too.

Page file

I have read countless articles about the page file, but I still haven't
figured a couple of things. I believe that a page file is needed even when
you have a huge amount of RAM installed, but i am looking for more
information regarding this matter. For instance, would it be advisable to
just turn it off and see if your system behaves better without it? Or is it
100% certain that it is better to have a page file regardless of the amount
of RAM installed.

NTFS

Can anybody reply the following? When you move, delete, create, rename files
etc, the changes must be written to the MFT (assuming NTFS). Does Windows XP
cache any part of the MFT, does it group maybe such activities to accelerate
them, or does it have to access the MFT every single time? Again, a reliable
resource will be appreciated.

Thank you in advance.

Fleur de Coin
- http://www.fleur-de-coin.com/
Åëëçíéêüò óôñáôüò
- http://skopia.digitalrice.com/

 

[Guest]

The amount of time saved by defragmenting your pagefile is not going to be
anywhere close to the amount of time you've spent "reading countless
articles" on the subject.

Know when to say when.

 

[Plato]

Does anyone have any statistics on the performance gains? Personally, I
think that today's hard disk drives are so fast, that a difference between a
moderately fragmented drive and a defragmented one will be negligible. Has

Correct. Dont waste time driving down the wrong street.

 

[Harry Ohrn]

Alex Nichol has a page here http://aumha.org/win5/a/xpvm.htm

 

[SlowJet]

The real answer is that the page file doesn't fragment in XP.
But wo knows, maybe you could use DOS to do it. :O

SJ

 

[David Candy]

Page files aren't accessed like normal files and fragmentation isn't that important (from nil to extremely, very, minor).

But page files shouldn't be fragmented in many fragments. As they are accessed differently to normal files the disk locations are kept in memory. Each fragment requires extra memory to describe the location. With 10 fragments this is almost nil impact. With tens of thousands it is substantial. You are likely to be near the it's not worth it level of fragmentation.

 

[yandr]

My questions were stricktly hypothetical.

But since you mentioned it, I believe that "The amount of time saved by
defragmenting your pagefile is not going to be anywhere close to the amount
of time you've spent defragmenting"

Fleur de Coin
- http://www.fleur-de-coin.com/
Åëëçíéêüò óôñáôüò
- http://skopia.digitalrice.com/

 

[yandr]

Now I see that I haven't phrased my post correctly.

I don't mean defraging the page file, I mean defraging in general.

The phrase "I was wondering about defrag and the page files" should have
read "I was wondering about 2 completely unrelated issues"

 

[Alex Nichol]

I have read countless articles about the page file, but I still haven't
figured a couple of things. I believe that a page file is needed even when
you have a huge amount of RAM installed, but i am looking for more
information regarding this matter. For instance, would it be advisable to
just turn it off and see if your system behaves better without it? Or is it
100% certain that it is better to have a page file regardless of the amount
of RAM installed.

Well, if you haven't done so, read my page www.aumha.org/win5/a/xpvm.htm
(And no, it might seem better without, but you would just lock out a
large part of your RAM). With sensible settings there will be no
traffic on the page file anyway.

Can anybody reply the following? When you move, delete, create, rename files
etc, the changes must be written to the MFT (assuming NTFS). Does Windows XP
cache any part of the MFT, does it group maybe such activities to accelerate
them, or does it have to access the MFT every single time? Again, a reliable
resource will be appreciated.

It does some caching of the records for active files, but on this one I
can't give detail

 

[Alex Nichol]

Defrag

Does anyone have any statistics on the performance gains? Personally, I
think that today's hard disk drives are so fast, that a difference between a
moderately fragmented drive and a defragmented one will be negligible. Has
anybody found reliable information on the net concerning this matter?

What matters in drive access is not so much actual transfer, but seek
time. That is measured in milliseconds, so if a file is significantly
fragmented, reading it gets slowed down quite a lot. This is much less
important for the Page file where single clusters are being accessed
randomly anyway - even if there is any noticeable traffic on the file

 

[Fuzzy Logic]

My questions were stricktly hypothetical.

But since you mentioned it, I believe that "The amount of time saved by
defragmenting your pagefile is not going to be anywhere close to the
amount of time you've spent defragmenting"

This free utility will defrag your page file. If, like me, you have defined
a minimum size it's worth doing once and doesn't take very long.

http://www.sysinternals.com/ntw2k/freeware/pagedefrag.shtml

 

[Larry(LJL269)]

What matters in drive access is not so much actual transfer, but seek
time. That is measured in milliseconds, so if a file is significantly
fragmented, reading it gets slowed down quite a lot

Here's sample from specs of 4 yo HD:
Seek Times (Typical):
Track to track: ....................... 0.9 ms
Average: .............................. 9.0 ms
Maximum: .............................. <20.0 ms
& DATA TRANSFER RATE: ............. Up to 33.0 MBs

So in time it takes to do avg seek, we could have read
~330KB. Or in 1ms, we read ~8 4K clusters.

HTH-Larry

Any advise given is my attempt to show appreciation for all
the excellent help I've received here but I'm no MVP so it
may only apply NUGS (Normally, Usually, Generally, Sometimes