PROBLEM: Backup copy directory highly fragmented

[stevea]

SuperDuper is advertised and promoted as a solution for disk defragmentation.

However, DiskWarrior's graph reveals that a backup copy made from SuperDuper contains 32% directory fragmentation.

Copies made by other utilities or the Finder do not exhibit this high degree of directory fragmentation.

This contradicts some of the benefits promoted for SuperDuper.





SuperDuper 1.5.5 (v74)
OS X 10.3.9
________
Ecigarettes (http://vaporizer.org/forum/electronic-cigarettes/)

[dnanian]

We quite clearly mention, in the white paper, that we "roughly defragment your system files". We make no claim of directory structure defragmentation, and also suggest people buy iDefrag if fragmentation is a serious issue.

We do nothing weird when we copy files, walking them in "natural folder order" as returned from fts, so I don't see how you'd get different results with other tools... the files are copied sequentially, in order....

[stevea]

We do nothing weird when we copy files, walking them in "natural folder order" as returned from fts, so I don't see how you'd get different results with other tools... the files are copied sequentially, in order....

That's why I would expect little or no fragmentation in the directory... so it was very strange to see that problem with SuperDuper.

Perhaps you can use DiskWarrior to conduct some experiments?
________
LAMBORGHINI LM001 HISTORY (http://www.lamborghini-tech.com/wiki/Lamborghini_LM001)

[dnanian]

As this is just a general side benefit of the copying operation (as we said in the paper), it's not something we're likely to spend a ton of time on. But, we'll certainly re-check the copy logic.

What exact tools did you use that showed different results? Were they file-by-file?

[stevea]

What exact tools did you use that showed different results? Were they file-by-file?

Finder, Synchronize! Pro X, several others I can't recall right now.

Copying whole volumes to blank disks, copying groups of folders to blank disks, etc.

Also, copying to disks containing data already that have defrag'ed directories- the directories remain mostly defrag'ed.


In fact, the original disk we copied with SuperDuper had a directory that was only 3% fragmented (according to DiskWarrior). Copying with SuperDuper resulted in a copied disk with a directory that was 32% fragmented. That is not good!...
________
Airsoft guns (http://airsoft-shop.info/)

[dnanian]

OK. I don't know what they're doing differently, but when we get a chance we'll take a look. It could be they're walking the disk structure with Carbon, which is giving the files back in a different order.

As I said, we use the Unix-level calls to walk the hierarchy. That's the order we recreate in, exactly as given to us...

[sjk]

Thanks for reporting this, stevea. I've long noticed DiskWarrior reports severe directory fragmentation on SuperDuper! destination volumes and had wondered about it, too, but never got around to mentioning it. In fact, one volume I checked yesterday was at 32%. :)

Oh, I thought it might be a side effect of using Smart Update over time instead of Erase then Copy but hadn't actually tested that.

[brich]

I also see this level of directory fragmentation from DW graphing, but only after doing a full clone....not a Smart Update to an existing clone. I've used DW for years, and it has graphed anywhere from 18% to 32% of items out of place following a Mac OS new install (depending on OS version). As a result, I usually run DW about once every couple of weeks, both from the clone and then from the internal to 'fix' the clone after confirming that the clone boots fine. I formerly used CCC, and seem to recall similar results, but the latter's incremental backups took much longer than SD...a real SD advantage.

Also, I have not experienced any adverse effects from the fragmentation so far, including the first two weeks of running Tiger where a compatible DW version wasn't yet available.

[dnanian]

The lack of noticeable performance penalty for this type of fragmentation doesn't surprise me terribly, as the OS deals with the directory structure in a pretty efficient way...

[brich]

The lack of noticeable performance penalty for this type of fragmentation doesn't surprise me terribly, as the OS deals with the directory structure in a pretty efficient way...

Actually, I've used DW mostly to help ensure maximum stability (as preventive maintenance) as much or more than for a perceived performance boost. Not sure if that's totally valid or not.

[dnanian]

Although I haven't found "preventative" use of DW helps me much (my file system has been quite stable), there's nothing wrong with running it as you're doing. But I'm not sure it'll help much pre-problem.