Which Disk Defragmenter?

Thanks, I'm aware that direct quotations aren't allowed, as I was sure that your interpretation would probably be more than adequate.

 

I did give Puran a try recently too. Compared to PerfectDisk 2008, it was a breath of fresh air in terms of installation size and resource usage (4.64MB disk/2.93KB registry space compared to PD's 30.28MB disk/95.55KB registry usage) and memory usage was far lower (can't remember exact details now, but it did have only one process compared to PD's 3 or 4). It was significantly faster at completing a defrag also.

However, just like PerfectDisk, it didn't make any noticeable difference compared to Windows' own defragger, its optimisation feature (PIOZR) seems utterly useless - no files were moved away from the first (slower) section of disk and, like too many other developers, Puran seems to think that a shiny, licky, non-standard user interface is the way to go.

One particular failing of Puran is its ability to move large files - when I used it on a partition containing just a couple of 50+GB files (image backups) it took over 2 hours to move one file. To judge from the disk noise (and low memory usage) it was moving it one sector at a time which would have been highly inefficient. When gigabytes of RAM are available, it should use them to read large files in bigger chunks.

On UI grounds alone, I'd fail Puran. I despise non-standard interfaces since they invariably result in the program not working with theming software like WindowBlinds (which, in its pre-DRM days, was a perfectly good choice for providing all the UI candy most people could want). However the lack of noticeable performance benefits is probably a greater issue for most users.

 

Good comments, even though you have given me a quesy feeling about Puran -- which I was thinking of buying. Also, I rather liked Puran's UI -- non-standard stuff helps me to feel "unique" (just like everybody else).

So... do you have an alternative to Puran which posses Puran's positives (at least most of them) and which lacks Puran's negatives (at least most of them)?

 

At the moment I can't see any reason to use anything other than Windows' own defrag utility. The only ways third-party software can improve on this is by (a) defragging data it can't (which either tends to be one-off tasks like the page/hibernation files or readily-cached data like MFTs) or (b) improving file placement.

Now the more I think about (b), the harder it seems to achieve in practice. Moving files to the fastest (outer) section of disk isn't hard but selecting files that benefit most (larger files to minimise seek time overheads, frequently accessed files to maximise read time savings) is trickier. Some tactics like Puran's gap filling may actually increase fragmentation for files regularly written to (log files being the best example) as such files would benefit more from having empty sections left after them. NTFS did initially do this, by providing "guard space" after every file in order to reduce subsequent fragmentation, but when disks reached 80-90% capacity, the only free space left was guard space causing fragmentation to skyrocket.

On top of this, you have the impact of security software (anti-virus scanners in particular) which complicate matters by scanning (and hence needing to read) their database every time a file is accessed and hard disks themselves which can dynamically re-assign sectors if one turns faulty - meaning that what a defragmentation utility thinks is adjacent data may well be at a different location on disk.

So third party defragmentation utilities are going to have a hard time justifying themselves and the increasing use of flash drives (where data location is of almost no consequence and highly variable due to such drives' write-levelling algorithms) is likely to be their death knell.

Now I don't think Puran is bad as such (certainly it performs better than PerfectDisk and most of the other high-profile utilities) but I could not, from my trial, see any great justification for continuing to use it. That doesn't mean that others can't draw a different conclusion.

I did rather like its block display though - if Puran turned that into a Breakout style game they'd be onto a winner.

 

Where can I find the discount info?

 

How do you know this? (Just curious)

More to the point, however -- On Puran's website I cannot find any statement whereat they say that they place "last accessed files" 2nd.

Instead. on their web page HERE, under the blurb for PIOZR, they say...

"Frequently used files" (Puran's statement) is NOT the same as "last accessed files" (your statement), is it?

Further, I am puzzled as to HOW Puran would determine frequency of file use unless it monitors usage, as you said that Ontrack used to do. I would appreciate your thoughts on this.

 

Here: http://www.puransoftware.com/Puran-Defrag-Order.html

Normal price = 25
Current price = 20

 

Here's what Puran support had to say:

 

As a small point of courtesy, if Puran wish to respond to issues raised here they should do so themselves - not via a third party. Copying replies on their behalf will always raise the issue of authenticity, regardless of intention and it is clear from those replies that they have not read my comments in full (i.e. I most certainly did trial Puran for 30 days). As such, please advise them to post directly if they wish to participate further.

To sum up the responses quickly though:

• Disk space is allocated from the centre outwards - this means that the last section of disk is the fastest, not the first. Defragmentation utilities that move data to the beginning of the disk therefore slow things down rather than speed them up (or they would, if Windows itself didn't default to using the starting tracks). So for Puran (or any other utility) to use the fastest areas of the disk, it would have to move files to the end, not the beginning. Raxco did move some files to the middle, Puran moved everything to the beginning (and it was run several times during that 30-day period).
• (Edit: ) Puran's statement "Puran Defrag does move frequently used files" begs the question, how does it know which are frequently used? It does not install a driver to monitor file system access (Raxco does install a driver but whether it does any such monitoring, I can't tell) so all it can go on are file attributes (like Last Accessed Time, which isn't an especially good guide) or the contents of XP's PreFetch cache (which covers programs only). It does add a .dll to Windows' System32 folder but this is not injected into other processes so it can't be using that for monitoring either.
• Taking 2 hours to move 50GB+ is most certainly not reasonable when the volume in question is a 4-drive RAID-0 array capable of 280MB/s sequential read and 250MB/s sequential write speeds (benchmarked using Sandra). A 50GB file should take about 6 minutes (380 seconds, 180 to read, 200 to write) to move at best speed. It was clear from the disk activity that Puran was moving this file by single sectors, crippling performance as a result.
• UI - I said it looked non-standard and it does. Simplicity isn't hard to achieve with a defragmenter so having a simple interface isn't anything special. It does however stick out like a sore thumb when compared with "standard applications" and I personally resent having to deal with such software. Not least, when you have several applications open all "doing their own thing" interface-wise you end up with a dog's breakfast for a desktop. This however is just a personal opinion and applies to most other defragmenters also.

 

Using UD for C: Windows Xp and programs set to 45 days past data was used for performance and 46 days for achive works perfectly for me. I does help that I don't have AV scanners messing up dates and I use Imaging to back up rather than traditional backup programs which often play hell with "days past data used". Very few files in the acrhive ever move with almost all activity ocurring at the outer edge of the drive.

 

I heartily agree. Evidently Puran disdains personal participation in this forum. Further, I would have expected their rejoinder to place greater emphasis on technical details, rather than being primarily anecdotal.

In any event it would be valuable to see *before & after* screenies of a major defrag done by Puran.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Having somewhat lost kokua for Puran, I am now considering UltimateDefrag, which has a paid version & a free version. Does anyone know the comparative differences between the 2 versions?

 

MST Defrag 3.0 is another defragger that has just now drawn my interest because of: (1) some favorable comments I have heard, (2) a price of only $15.90 for home users, & (3) their prompt reply to email.

Speaking of email, I sent them one with 3 questions. My questions, & the substance of their answers, are given below...

1- Will MST defrag at boot-up? Yes, MST Defrag includes a Boot time defragmentation component.

2- Does it defrag MFT? Yes

3- Does it use any optimization algorithm to improve file placement & speed up HD access times? Yes, it supports 3 different defragmentation algorithms.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Update- I downloaded MST for trial. Trial is limited to 15 days, which is okay. However, unless I did something wrong, the trial version is crippleware to the extreme.

Therefore I cannot really tell very much about it except that it has a boring GUI & 90% of everything it supposedly does is grayed out.

 

Tried MST Defrag and could see no way to control file location. any defrag program can defrag what is needed is the ability to move unused and rarely used files out of the way. small active file defraging then makes more sense.

 

What features are you not able to access? When I demoed it, everything worked. Did you dl the home version?

 

Basically everything was grayed out except simple defrag.

Yes -- the home version. And I put "x" in the "trial" box during install. If it worked for you then I undoubtedly did something wrong during install. I shall give it another trial one of these days.

In the meantime, I am very impressed by the free version of UD. Once I studied the beautifully written Help pdf, I was able to configure it to really speed things up.

 

Would someone help straighten this out please?

OK, any circle, and this case the metallic hard drive platter has an "outer" rim and an "inner" rim. It's logical to me that anything within the inner rim travels at the highest velocity compared to the outer rim which is farther from central core. I been told that both rim areas travel at the same rate because of the principle of binding together from being a single object/ molecular entity.
OK i suppose, but for the sake of this device known as a PC's Hard Drive, it would seem to me that the best performance of data acquisition (seek/read) due to the separate structure known as the "heads" accessing data on this platter, would be the inner orbit not the outer orbit as many suggest.

Or maybe my physics isn't matching my Logic when it comes to this particular device

ANYONE?

THANKS

 

Hi EASTER
1. Outer rim is "longer" than inner rim
2. So outer rim can have "longer row" of 1's & 0's
3. When the disc turns one full circle:
- the head when placed at inner rim will read 1 full "row" of 1's & 0's
- the head when placed at outer rim will also read 1 full "row" of 1's & 0's
- but since there's a "longer row" of 1's & 0's at outer rim, more data is read/written in 1 full turn -> higher rate of read/write, better performance at outer rim

Hope my very crude explanation makes sense to you.

 

Yeow very good at least i get it,but there's more then only outer versus innerrim,files has to be contiguous and really most used files at the first outer tracks(frequent used app. and system),anything else behind this,i believe that there is no need to place least used files on the inner tracks because of the system search algorithm(everything in one place at the outside of disk)prevent wear and tear on platter and read/write heads.The benefit of dividing to place files on outer and inner tracks is minimal rather then placing everything on the outer rims which is best performance IMO.I must say though that with modern Sata and sure the new raptors and for really sure the coming SSD drives benefits of manually editing the fileplacement are diminishing to a point that its not needed anymore.

 

For SSD defrag is not only not necessary but actually reduces the life of the drive. At the moment I think that the best non raid solution is a small SSD (30 gb) and a larger and fast data drive. UD2008 is still the best solution for the data drive

 

How often do you read a file end to end on the OS in a non-streaming (streaming implies nowhere near performance limited) manner, my tests shows this is very rare case.

Where do you suggest placing a file that WAS in the way of a file that was to be placed on outer track ? Its gotta go somewhere.
Its accepted by most people that creating big contiguous areas of freespace is good for reducing future fragmentation, so placing files at the opposite end of the drive is the most logical place for increasing contiguous freespace.

What system search algorithm are you referring to?

Wear and tear ?
Magnets do not wear out, the heads do not touch the platter, if they do, they cause great damage and drive requires repair (usually).

If a file is less frequently used, where-ever it is placed will not impact performance in a significant manner compared to the placement of the most frequently uses files

I agree. Been a long time since the OS even knows about underlying physical drive structure, therefore any representation of the physical structure is guesswork.

In the case of Zoned Bit Recording, the OS has no clue about which logical tracks are more dense than the others, is it the first 1, 2 or 50 Does it increase gradually ?

Here's a useful snippet:

http://www.storagereview.com/guide2000/ref/hdd/geom/geomPhysical.html

 

I get what you're saying, but it's hard to read. For slow thinkers like me, it would help if you typed a little slower so that you don't run words & sentences together.

 

I still haven't heard a good explanation of how Disktrix (or any other defragger for that matter) correlates its disk map to the actual geometry of a multi-platter drive. I don't even know if the defrag API allows you to read/write on which actual platter a piece of data resides.

 

Even the BIOS only reports a representation of geometry to the OS, so the OS does not know the exact geometry.

One safe assumption that you can make though is that whatever numbering scheme is used, is that it is usually/mostly sequential though the drive, so the lowest figures (whatever is reported, LBA address, fake block/track/sector figure etc) is on the outer part of the disk and the higher figures are on the inner part of this disk.
Diskeeper is the only defragger I know of that does take figures across the drive so it can work out what is the fastest and slowest parts of the drive.

Heres a couple of hd tach outputs I googled:

http://www.ihaveacrazywife.com/images/HDTach320.jpg

http://images.anandtech.com/iblog/raptor150sb600hdtach301S.jpg

As you can see these figures for the Raptor and Barracuda, 2 popular drives show that to drives do sequentially slow according the the logical disk access.