Which defrag?

"Enkidu" wrote:

You should look at some studies on the effects of fragmentation:
http://www1.execsoft.com/pdf/Diskeeper_Evaluation.pdf

Hmm, do the words "vested interest" mean anything to you? I
notice that they are very careful to leave out any mention
of disk caching, paging, running from memory and all the
other things that are done today to speed up applications.

If an application does a read, process, write, read,
process, write cycle all the time there might be benefits
from defragging. Typical applications don't do that.

Only if an application is heavily I/O bound is there any
benefit from careful placement of files on disks. The only
real-world example I can think of is backup and from my
tests the benefits were only a few percent.

Although I disagree with you about defragmentation, you are actually closer
to the truth than many people might think, especially if you have a machine
with lots of RAM and you either leave your machine on all the time or you
merely log out without actually rebooting, i.e. a system in which lots of
program code and data end up in the RAM system cache and stay there for long
periods of time. The point is that -- other things being equal -- CPU and
RAM probably account for about 95-97% of system performance, with a
defragmented and junk-free hard drive accounting for the remaining 3-5
percent. There is no good justification for putting up with a 3-5 percent
performance hit by not regularly defragging doing idle periods, but on the
other hand the benefits of defragmentation are often overhyped, especially by
the third party defragmentation vendors.

Ken

"da_test" wrote:

I'm a little dubious of it's "smart placement". Perfectdisk works
better that DK (faster) on Fat32.

I am, too. In particular, PerfectDisk moves the MFT file to about 1/3
inside the drive. When I asked them about this, they referred me to a
Microsoft Knowledge Base Article that suggested that this placement optimized
performance. I read the actual article, and it actually said 3-5 GB inside
the drive, not 1/3 inside the drive. The difference between 3-5 GB and 1/3
inside the drive is huge, especially if you have a very large HD, as I do
(250 GB).

I also asked them about whether placing least modified files at the
beginning of the drive made sense from a performance standpoint, other than
the point that this setup minimizes refragmentation. I no longer remember
the exact explanation I received, but I do remember it making good sense at
the time. In any event, any performance drop by putting, say, an old but
very large music file closer to the outisde of the disk will probably be too
small to notice anyway. Even so, I like the way the native defragger puts
these behemoth files in a separate part of the drive from smaller files.
[Note: I think Diskeeper will also move these huge files closer to the
outside of the drive, while the native defragger leaves them further away.
In this regard, if I get a choice, I prefer the behavior of the native
defragger.]

Incidentally, I have found that both Raxco (PerfectDisk) and Executive
Software (Diskeeper) have excellent customer service and technical support.
One can learn all sorts of cool stuff about hard drives and fragmentation by
e-mailing them questions.

Ken

That's also part of the philosophy behind PerfectDisk, and PerfectDisk
otherwise does a much better job than Norton.

"Dave" wrote:

I have used two, Diskeeper and Norton's Speed Disk which use to come
with Norton's Utilities. I stayed with Speed Disk because it does
something that Diskeeper doesn't... during defragging, it fills up
all the empty spaces left which seems (to me anyway) to cut down on
how quickly the whole thing gets fragmented again.

Regards,
DW

"jt" wrote in message . ..

Odd. I downloaded and ran the trial version of O&O and it flew on my
machine. By contrast, the trial of PerfectDisk crawled and after it was
done, my system was slow as mud. I then ran O&O again and it was much
slower this time, but after it ran my system was at least fast again. I've
seen many recommendations for PerfectDisk, but after my experience I
wouldn't buy it with my neighbor's money.

I recently played around with PerfectDisk and O&O Defrag. The target
was a 70% full 4.5 year old 20GB drive that has never been reloaded
with OS, but that has seen alot of use, particularly when it comes to
application installs/removes updates. FWIW, the drive has been regularly
defragged using the built-in. My goal was to see what each defragger
could do if time were of no issue, so I didn't play around with some of
the options. I did a boot defrag using PerfectDisk and it took a fairly
long time. There was little if any improvement in boot time, but the system
did feel a tiny bit more snappy. Later I did a complete by name type
defrag using O&O, and IIRC it took considerably longer to do its thing.
Again, there was little if any change in boot time. However, there was a
noticeable improvement in application load times, presumably due to
the restoration of application files proximity.

I need to spend some more time evaluating the defraggers, but I must
say that I do like the concept of being able to restore proximity. It
would be nice if users had even more control over where dirs/files
get placed (I'd experiment with putting archived and infrequently used
stuff at the highest LCNs).

On Mon, 28 Mar 2005 16:13:03 -0800, "Ken Gardner"
wrote:

That's also part of the philosophy behind PerfectDisk, and PerfectDisk
otherwise does a much better job than Norton.

Well, I took the plunge and downloaded, installed and ran a trial copy
of PerfectDisk. Must say that I like it. However, I do wish it would
allow you to configure how you would like to sort your
files/directories. An included PDF manual would also be nice.
However, the defrag on boot is nice. I can see that you get a better
defragging and "hole filling" (some call it compressing) while at the
system level. I also like the fact that they don't treat you like a
common criminal by making you prove that you didn't steal their
software after paying for it by using that activation trash certain
companies are adopting. Being so, it will be more than a pleasure to
send them money to register their software. Its nice to be treated
with respect instead of being treated with suspicion of thievery like
you are by the activation scheme crowd.

Regards,
DW

Ken Gardner wrote:
"Enkidu" wrote:


You should look at some studies on the effects of fragmentation:
http://www1.execsoft.com/pdf/Diskeeper_Evaluation.pdf

Hmm, do the words "vested interest" mean anything to you? I
notice that they are very careful to leave out any mention
of disk caching, paging, running from memory and all the
other things that are done today to speed up applications.

If an application does a read, process, write, read,
process, write cycle all the time there might be benefits
from defragging. Typical applications don't do that.

Only if an application is heavily I/O bound is there any
benefit from careful placement of files on disks. The only
real-world example I can think of is backup and from my
tests the benefits were only a few percent.


Although I disagree with you about defragmentation, you are
actually closer to the truth than many people might think,
especially if you have a machine with lots of RAM and you
either leave your machine on all the time or you merely log
out without actually rebooting, i.e. a system in which
lots of
program code and data end up in the RAM system cache and stay
there for long periods of time.

Yes, all the testing that I've done was on servers, which
typically have the characteristics you mention.

The point is that -- other things being equal -- CPU and
RAM probably account for about 95-97% of system performance,
with a defragmented and junk-free hard drive accounting for
the remaining 3-5 percent.

If you mean that you could improve performance by those
%ages, I'd agree with you. Disk is so slow relatively,
though, that the 3-5% improvement *should* make a huge
difference. In a server situation though, I've not seen it
make a noticeable difference.

One test I'd like to make, if I were doing this today would
be to run a script-based testing tool rather than use simple
batch files.

There is no good justification for putting up with a 3-5
percent performance hit by not regularly defragging doing
idle periods, but on the other hand the benefits of
defragmentation are often overhyped, especially by
the third party defragmentation vendors.

There very few independant studies. Search for
defragmentation on the web and you there is almost no hard
evidence. There's plenty of references to Microsoft
documents which talk about defragmentation and how it works,
but little evidence that quantifies the possible
improvements. Which are likely to be different for database
servers or web servers or workstations or home machines.

There was a need for defragmentation back in the early days
of Windows with small, slow disks on FAT16 filesystems. I'm
not convinced there's a need when we have large, fast disks
and NTFS filesystems. Not to mention large amounts of RAM.

Cheers,

Cliff

--

Barzoomian the Martian - http://barzoomian.blogspot.com

So, now that you've taken the plunge, how much faster does your system run?
How much faster does it boot? Got any before and after benchmarks to share?

Modem Ani

"Dave" wrote in message
...
On Mon, 28 Mar 2005 16:13:03 -0800, "Ken Gardner"
wrote:

That's also part of the philosophy behind PerfectDisk, and PerfectDisk
otherwise does a much better job than Norton.

Well, I took the plunge and downloaded, installed and ran a trial copy
of PerfectDisk. Must say that I like it. However, I do wish it would
allow you to configure how you would like to sort your
files/directories. An included PDF manual would also be nice.
However, the defrag on boot is nice. I can see that you get a better
defragging and "hole filling" (some call it compressing) while at the
system level. I also like the fact that they don't treat you like a
common criminal by making you prove that you didn't steal their
software after paying for it by using that activation trash certain
companies are adopting. Being so, it will be more than a pleasure to
send them money to register their software. Its nice to be treated
with respect instead of being treated with suspicion of thievery like
you are by the activation scheme crowd.

Regards,
DW

"Modem Ani" wrote:

So, now that you've taken the plunge, how much faster does your system run?
How much faster does it boot? Got any before and after benchmarks to share?

If only people would answer these types of questions after trying all
so-called "performance enhancing" software, they would have a much better
understanding of how well XP actually performs even when you don't mess with
it beyond what it is already designed to do.

Ken

"Enkidu" wrote:

There very few independant studies. Search for
defragmentation on the web and you there is almost no hard
evidence. There's plenty of references to Microsoft
documents which talk about defragmentation and how it works,
but little evidence that quantifies the possible
improvements. Which are likely to be different for database
servers or web servers or workstations or home machines.

These points are sources of constant frustration that I have with most
so-called performance enhancing software, including but not limited to
defraggers. Intellectually I am perfectly capable of understanding that
other things being equal, a defragmented hard drive will out-perform a
fragmented hard drive. But are we talking here about seconds, tenths of
seconds, or milliseconds?

My own personal experience, which is as a workstation user, is that a
regularly defragmented hard drive can save you seconds rather than
milliseconds in disk drive operations, i.e. you can actually notice the
difference. However, I cannot notice any transparent difference between
defragging a hard drive using the XP built-in defragger and defragging using
a third party program such as Diskeeper or PerfectDisk. Of these programs,
only Diskeeper even attempts to measure the performance improvement you might
gain, but it does so in terms of percentages rather than actual time. If it
takes ten milliseconds to load a file when it used to take five milliseconds,
that may be a 50% improvement but no human being will ever notice it. If,
instead, we are taking about tenths of seconds, then the improvement will be
noticable.

Ken




There was a need for defragmentation back in the early days
of Windows with small, slow disks on FAT16 filesystems. I'm
not convinced there's a need when we have large, fast disks
and NTFS filesystems. Not to mention large amounts of RAM.

Cheers,

Cliff

--

Barzoomian the Martian - http://barzoomian.blogspot.com

"My own personal experience, which is as a workstation user, is that a
regularly defragmented hard drive can save you seconds rather than
milliseconds in disk drive operations, i.e. you can actually notice the
difference."

I think most people would agree with you.

Many users do not seem to realize that XP performs partial defrags in the
background, and that the design of these defrags - as I understand it - was
well thought-out to get the best bang for the buck. A third party defragger
is 'improving' on regular partial defrags, not on a system that has not been
defragged at all for months.

I have no problem if someone wants to use a third party defragger. If it
makes them feel better about their system and does no harm, why not go for
it. For me, the unmeasurable improvement in performance is not worth the
extra lines of code in RAM or the extra CPU. In my experience, a leaner
configuration runs best.

Modem Ani


"Ken Gardner" wrote in message
...
"Enkidu" wrote:

There very few independant studies. Search for
defragmentation on the web and you there is almost no hard
evidence. There's plenty of references to Microsoft
documents which talk about defragmentation and how it works,
but little evidence that quantifies the possible
improvements. Which are likely to be different for database
servers or web servers or workstations or home machines.

These points are sources of constant frustration that I have with most
so-called performance enhancing software, including but not limited to
defraggers. Intellectually I am perfectly capable of understanding that
other things being equal, a defragmented hard drive will out-perform a
fragmented hard drive. But are we talking here about seconds, tenths of
seconds, or milliseconds?

My own personal experience, which is as a workstation user, is that a
regularly defragmented hard drive can save you seconds rather than
milliseconds in disk drive operations, i.e. you can actually notice the
difference. However, I cannot notice any transparent difference between
defragging a hard drive using the XP built-in defragger and defragging
using
a third party program such as Diskeeper or PerfectDisk. Of these
programs,
only Diskeeper even attempts to measure the performance improvement you
might
gain, but it does so in terms of percentages rather than actual time. If
it
takes ten milliseconds to load a file when it used to take five
milliseconds,
that may be a 50% improvement but no human being will ever notice it. If,
instead, we are taking about tenths of seconds, then the improvement will
be
noticable.

Ken




There was a need for defragmentation back in the early days
of Windows with small, slow disks on FAT16 filesystems. I'm
not convinced there's a need when we have large, fast disks
and NTFS filesystems. Not to mention large amounts of RAM.

Cheers,

Cliff

--

Barzoomian the Martian - http://barzoomian.blogspot.com

Modem Ani wrote:

Many users do not seem to realize that XP performs
partial defrags in the background, and that the design
of these defrags - as I understand it - was well
thought-out to get the best bang for the buck.

That's interesting! Got a reference for that?

Cheers,

Cliff

--

Barzoomian the Martian - http://barzoomian.blogspot.com

Are you looking for a reference about whether XP performs partial defrags in
the background or whether these partial defrags were designed with "best
bang for the buck"?

If you're asking about partial defragmentation, information on this abounds.
For example, this from TechNet: "Once every three days, by default, Windows
XP will perform a partial defragmentation and adjust the layout of the disk
based upon current use."
http://www.microsoft.com/technet/pro...te/xpperf.mspx

If you're asking about "best bang for the buck" - sorry, while I have read
that more than once I can't remember a specific reference right now.

Modem Ani

"Enkidu" wrote in message
...
Modem Ani wrote:

Many users do not seem to realize that XP performs
partial defrags in the background, and that the design
of these defrags - as I understand it - was well
thought-out to get the best bang for the buck.

That's interesting! Got a reference for that?

Cheers,

Cliff

--

Barzoomian the Martian - http://barzoomian.blogspot.com

On Tue, 29 Mar 2005 17:26:04 -0500, "Modem Ani"
wrote:

Your source only recommends 1 defragmentation and that's only after the OS is
installed.

Defrag : Therefore, it is a good idea to defragment the disk following an installation.


Are you looking for a reference about whether XP performs partial defrags in
the background or whether these partial defrags were designed with "best
bang for the buck"?

If you're asking about partial defragmentation, information on this abounds.
For example, this from TechNet: "Once every three days, by default, Windows
XP will perform a partial defragmentation and adjust the layout of the disk
based upon current use."
http://www.microsoft.com/technet/pro...te/xpperf.mspx

If you're asking about "best bang for the buck" - sorry, while I have read
that more than once I can't remember a specific reference right now.

Modem Ani

"Enkidu" wrote in message
...
Modem Ani wrote:

Many users do not seem to realize that XP performs
partial defrags in the background, and that the design
of these defrags - as I understand it - was well
thought-out to get the best bang for the buck.

That's interesting! Got a reference for that?

Cheers,

Cliff

--

Barzoomian the Martian - http://barzoomian.blogspot.com


--
more pix @ http://members.toast.net/cbminfo/index.html

On Tue, 29 Mar 2005 09:34:08 -0500, "Modem Ani"
wrote:

So, now that you've taken the plunge, how much faster does your system run?
How much faster does it boot? Got any before and after benchmarks to share?

Well, as my past message indicates, I have only used it for a day now
so lets see if I can tell you what I have seen in one day.

And, as my previous posts indicated, I was using Norton's Speed Disk
which works much the same way as PerfectDisk in that it not only
defrags but moves everything forward to fill the "holes" which means
that refragmentation takes a bit longer to occur unlike on systems
that were only defragmented but the bunches of open "holes" were left
to excite further fragmentation.

So, with that, I have seen no difference between any speed taken to
load the mega CAD files we load on and off all day long between
Norton's Speed Disk and PerfectDisk. However, the defrag utilities
that I had tried before going to Norton's speed disk and now testing
PerfectDisk did show that they didn't do much to keep the loading time
on these mega CAD files down. I found out that the reason was because
they left open holes and every time a mega sized CAD file was saved,
it just starting filling in all the holes and thus became fragmented
and took longer to load up next time around.

So, All I can say is that in the specific type of operations we do
with computers, defragmenting Along With compression or filling the
holes up or whatever the specific utility calls it, does better at
mega CAD file loading performance. I see no sense in defragmenting if
you aren't going to fill up all the holes left behind. Sorry if I
don't use the proper technical terms for everything but I'm just a
business owner that looks for anything to keep the performance of our
CAD machines up to snuff. I go with what my operators tell me works
because they are on the systems 8-10 hours/day. They say
Defragmenting WITH Compression (filling the holes up) is the ticket.

Even though Speed Disk and PerfectDisk operate the same (give the same
performance benefits), I like this PerfectDisk's GUI and graphical
progress indicator better that Norton's. That is why I am leaning
toward changing from Norton to PerfectDisk and of course as long as it
keeps my CAD operators happy.

Hope this answered your question.

Regards,
DW

On Tue, 29 Mar 2005 08:51:04 -0800, "Ken Gardner"
wrote:

If only people would answer these types of questions after trying all
so-called "performance enhancing" software, they would have a much better
understanding of how well XP actually performs even when you don't mess with
it beyond what it is already designed to do.

If you want people to "answer these types of questions" then have the
common courtesy to allow them more than a few hours to evaluate the
software......

DW