how severe is bug 941715, and can I fix it?

[captainmidnight]

I am tired of using windows' builtin Disk Defragmenter as it is so slow (among other things, it executes in really low priority, which is not what I want as I like my defraggers run in a quick mode, at high priority, every night as part of a batch process that includes things like backup and free space wiping).

My web search turned up JkDefrag as the most promising open source windows defragger--thank you jeroen!

But under the known issues on the home page is a mention of a serious sounding MS bug:
    http://support.microsoft.com/kb/941715/en

As all my drives are SATA and I am running win xp sp2 home, I am worried about running any defragger until I know that this bug won't corrupt my data.

Am I overworried here, or should I wait until this bug is fixed?

I looked at that MS webpage, but there is no link to download the hotfix.  I then clicked thru several of their support webapges until arriving here
    http://support.microsoft.com/oas/default.aspx?ln=en-us&prid=3219&gprid=37013
where it appears that I am unable to even email them about this hotfix unless I have some sort of support contract; they do not seem to respond to emails from their general users.  Or did I miss something?  Is there some other way that I can get this hotfix?

I do apply the monthly MS patch updates everytime that they come out, as soon as windows tells me that they are available.  But for reasons known only to them, MS has chosen not to release this hotfix yet as part of their standard monthly updates.

By the way, how much of a speedup should I expect to see with JkDefrag over the standard windows defragger?  I intend to run it everyday as mentioned above, so you can give me the steady state answer as opposed to how long JkDefrag takes to execute the first time.

[jeroen]

Am I overworried here, or should I wait until this bug is fixed?

The users that alerted me to this problem have also said that the files that got corrupted were system files only, if I remember correctly, and could be repaired by running chkdsk. The Microsoft bugreport does not mention the Microsoft defragmenter, which is strange because the FSCTL_MOVE_FILE system call was specially developed just for that defragmenter. And from the bugreport I gather the problem can only occur if the computer is shut down in the middle of a defragment operation. Put it all together and I think you can safely use JkDefrag. Since it is a problem inside the NTFS drivers there is nothing I can do about it.

Is there some other way that I can get this hotfix?

The Microsoft bugreport mentions a file size and version of the "Ntfs.sys" file. I gather that your file is different? Perhaps somebody who has the new file can post it.

how much of a speedup should I expect to see with JkDefrag over the standard windows defragger?

Results vary by how the computer is used, of course. The biggest performace benefit is gained by placing less important files in the SpaceHogs zone. This causes the more important files to be closer together, and also closer to the fast end of the disk. I am not going to boast with inflated percentages such as the commercial defragmenters do, but most users are surprised by how much faster their system becomes by using JkDefrag.

[captainmidnight]

The users that alerted me to this problem have also said that the files that got corrupted were system files only, if I remember correctly, and could be repaired by running chkdsk. The Microsoft bugreport does not mention the Microsoft defragmenter, which is strange because the FSCTL_MOVE_FILE system call was specially developed just for that defragmenter. And from the bugreport I gather the problem can only occur if the computer is shut down in the middle of a defragment operation. Put it all together and I think you can safely use JkDefrag.

I will install it tomorrow then.

The Microsoft bugreport mentions a file size and version of the "Ntfs.sys" file. I gather that your file is different?

Right: I looked at the file version of Ntfs.sys reported on the hotfix and compared it my local version number, and saw that mine was a much smaller number, and so assume that I have an old version.

how much of a speedup should I expect to see with JkDefrag over the standard windows defragger?

Results vary by how the computer is used, of course. The biggest performace benefit is gained by placing less important files in the SpaceHogs zone. This causes the more important files to be closer together, and also closer to the fast end of the disk. I am not going to boast with inflated percentages such as the commercial defragmenters do, but most users are surprised by how much faster their system becomes by using JkDefrag.

Sorry that I was unclear: what I was asking about was not the speedup of my machine during file operations as a result of defragging, but rather, is JkDefrag faster at doing th defrag operation itself compared to windows defragger?

Jeroen: thanks for your prompt response.

[jeroen]

is JkDefrag faster at doing th defrag operation itself compared to windows defragger?

I have never measured it. JkDefrag uses the exact same system calls, and the speed of the disk is the same, so defragmenting a particular file should take exactly as much time. There may be differences in total time, though, caused by such things as time needed to analyze the disk, and where a defragmented file is moved to.

[captainmidnight]

is JkDefrag faster at doing th defrag operation itself compared to windows defragger?

I have never measured it. JkDefrag uses the exact same system calls, and the speed of the disk is the same, so defragmenting a particular file should take exactly as much time. There may be differences in total time, though, caused by such things as time needed to analyze the disk, and where a defragmented file is moved to.

I have some -preliminary- results to share.

Here are several disk defrag times (in minutes) when last used MS's defrag.exe program:
   C: 34
   D: 1
   F: 9
   G: 80

And, for comparison, here are the corresponding disk defrag times the last TWO times that used JkDefrag:
   C: 14, 19
   D: 1, 1
   F: 1, 3
   G: 110, 110

The preliminary conclusion seems clear: with 3 out of the 4 disks JkDefrag was MUCH (2-3X) faster than defrag.exe; I am not sure what the deal is with disk G.  Congratulations Jeroen; I am sure that this is news that you want to hear.

Unfortunately, before you break out the champagne, these results have the following issues:
   1) limited number of measurements
   2) each time I benchmarked the programs, they faced moderately different disk conditions, so the comparisons are not consistent

Therefore, do not put much weight in these measurements until more people report more results.  Which is something that I hope other people on this forum will do...

As for why JkDefrag is faster than defrag.exe when both use the same api, my guess is that unless you tell it to run at a lower priority (via the -s option), it runs at full speed.  Maybe defrag.exe is written to run at lower priority or something.

[jeroen]

with 3 out of the 4 disks JkDefrag was MUCH (2-3X) faster than defrag.exe

Thanks for measuring! Did you measure with JkDefrag's "-a 2" option? Or the default "-a 3"? The build-in Windows defragmenter only defragments, it does not optimize.

[Esprit]

with 3 out of the 4 disks JkDefrag was MUCH (2-3X) faster than defrag.exe

Thanks for measuring! Did you measure with JkDefrag's "-a 2" option? Or the default "-a 3"? The build-in Windows defragmenter only defragments, it does not optimize.

Windows defragmenter in Windows XP 32 SP2 optimizes disc but not always. Often when you run it for the first time it only defragments what is capable. When you run it and it will see that most files are contiguous then it will do something like force fill.

[Esprit]

is JkDefrag faster at doing th defrag operation itself compared to windows defragger?

I have never measured it. JkDefrag uses the exact same system calls, and the speed of the disk is the same, so defragmenting a particular file should take exactly as much time. There may be differences in total time, though, caused by such things as time needed to analyze the disk, and where a defragmented file is moved to.

Moving of particular fragment takes the same time but on my PC is that dfrgntfs.exe uses half of memory than JkDefrag and CPU usage is about 25 % when JkDefrag has about 75 % CPU load. So the speed of defrag is based also on CPU speed! On the other hand JkDefrag uses more sophisticated algorithms so it can be faster or more efficient.

[captainmidnight]

with 3 out of the 4 disks JkDefrag was MUCH (2-3X) faster than defrag.exe

Thanks for measuring! Did you measure with JkDefrag's "-a 2" option? Or the default "-a 3"? The build-in Windows defragmenter only defragments, it does not optimize.

Sorry for not mentioning that: I use "-a 2".

My complete command line for a disk like C is
   

Code:

JkDefragCmd.exe -a 2 -d 2 C:\