Defragging SSD?

I have a 128GB SSD in my M1530. I was running a defrag this morning, and I started wondering how necessary it really is with this technology. Since there is not a head to move around to access data, I can't help but wonder if it is really necessary at all. In most of the speed results I have seen, SSD is the fastest (compared to a hard drive) when accessing random data from the drive. In fact, with modern caching a hard drive accessing data in a sequential fashion can actually be faster than SSD.

So I wonder.... why run a defrag at all? I'm guessing it's just a waste of time.

Thoughts?

-jk

 

All I know is that defragging a SSD drive causes more bad than good, it'll shorten the life span of the drive. Seek times are already negligible, so no need to worry about the defragging. Todays file systems aren't exactly optimized for a SSD drive.

I never owned one myself so I'm not that much into the details, I'll be getting one soon, though.

Edit,
Here you got a video talking about the bads of SSD drives and their fragmentation:
http://defragtv.wordpress.com/2008/08/22/defragmenting-ssd-drives-good-or-bad/

 

The above is correct. SSDs should NOT ever be defragmented...

 

Ahhh where is Les, where is Les??? The best out here to answer your query once n for ever..

and yeah, SSD' should not be defragmented, as far as I learnt from reading around.. I don't use a SSD myself though..

 

Here, have you seen these threads????? Just in case you didn't

http://forum.notebookreview.com/showthread.php?t=176076

http://forum.notebookreview.com/showthread.php?p=3036991#post3036991

http://forum.notebookreview.com/showthread.php?p=2871823#post2871823

http://forum.notebookreview.com/showthread.php?t=183474

 

I've only had to degrag...once.....and am trying to keep it that way

And, Les's guides are simply awesome


Cin

 

Ahhh.. never knew you were with a SSD... nice..

 

Wow...we get search back again and I get to see who is talking about me. Hey Cin! Nice to see the pik.

Defragging issue...is an old sell. An ssd has something called "wear levelling algorythms" which negates the need completely to defrag your ssd. In laments terms, the controller of the ssd constantly monitors all cells for predicted life span and moves information around without your knowledge to ensure all cells wear equally. This, in essence, is the ssd version of defragmenting a drive which is done automatically.

Similarly, information is not picked up the same on a ssd as it is on a hard drive. Much of the fragmentation of the hard drive is caused by the high speed of the spinning disk accompanied by the arm trying to pickup and place info.

Does this help?

EDIT: Thanks again all for the support....

 

And, he appears! WoOT....
OH...thanks for the compl.

Yes, the info helps me! Thanks, Les
(your guides are always solid BTW)


Cin

 

Attack of the smileys!
Save yourselves!

 

HA!!!!! You can run w/in NBR...but you cannot hide from the attack my smileys ~ they are special!!!!! WoOT....

:yes: :laugh: :tongue: :wink: :rolleyes2:

^^^^^Special for you... lmao.

Now back on topic!

Cin.

 

Thanks for that LES As always, crisp n neat n clear Cheers... nice to see u around

 

Les is truly the resident knowledgebase for SSD:s at NBR and his advice is correct "Do not defragment SSDs". However, he seems to have mixed up a few concepts here.

First, I think Les meant latency, rather than fragmentation below

Files (and directories of files) get fragmented (spread out in many pieces) on a hard drive because files are constantly being deleted, updated and also new files are being added to a hard drive during normal use. The negative effect of this can be that one single file can be spread out on many different locations on a hard disk. When the computer needs to access the file, the drive head needs to move to multiple locations on the rotating disk and wait for the actual file fragment to arrive during a disk revolution.

Fragmented files on hard drives thus result in additional latency (delays) when needed to be accessed.

Defragmentation of hard drives fixes this by consolidating file fragments (and directories of files) in sequential locations on the hard disk. Defragmentation of hard drives can also increase system speed by placing system files at the beginning of the hard disk (the outer edge of the disk), where the hard drive transfer speed is the highest.

Wear-leveling moves data around in order to even out the write usage of each memory block in an SSD. This is because flash memory only manages a limited number of writes, before failing. The wear-leveling algorithms on an SSD actually result in file fragmentation, which would be very bad for a hard disk.

However, an SSD is approx 100 times faster in accessing data than a hard disk. Also, in theory the access time to files on an SSD should be almost the same regardless of how fragmented the files are. So the file fragmentation of the data on an SSD due to wear-leveling has little or no impact on the performance. This means that "regular" fragmentation of files on an SSD also has little or no impact on performance.

On an endnote, while a hard disk actually shows how data is dispersed to a defragmentation utility, an SSD has an internal representation that does not correspond to the external one. Theoretically, your defragmentation utility could be showing that your SSD is 100% defragmented, while the internal state is completely different, due to the effects of wear-leveling.

 

I have to correct myself. It would seem that file fragmentation indeed does have a negative impact on SSD performance. However, it would also seem that the fragmentation needs to be extreme in order for the performance impact to be significant.

 

Glad you corrected yourself... and this is the first I have heard of this concept:

"The wear-leveling algorithms on an SSD actually result in file fragmentation, which would be very bad for a hard disk."

My concepts are not incorrect as it is a given that the latency or "time to access" a file is much slower. This I didn't believe need be explained as, after all, it is part of the whole ssd vs hd concept which all are aware. The magic of a ssd is the access time compared to a hd whether the hd is fragmented or not.

To close, I have been taught by many reputable sources that "wear levelling" is the ssds internal mechanism to not only ensure ssd life, but also, to ensure file defragmentation is not necessary.

 

thanks all, especially Les. I guess I won't be defragging again any time soon. I always found it a PITA anyway.

 

I occasionally check just to see if my system recommends defragmentation. In the year or so since I have been running ssd, I havent had it recommend that yet and, in fact, my system is still as fastas its ever been. I am running the Samsung SATA II presently and just received a new Memoright GT 64Gb which I may throw in and try out.

 

I am not sure that I expressed myself clearly and I am not sure I understand your reply (so much for my communication skills ;-). My correction was in regards to my incorrect statement, that SSD performance was unaffected by fragmentation.

Above you are confusing fragmentation with latency.

Unless the SSD itself is filesystem/filestructure/file aware (which it is not) it is impossible for it to defragment itself (in conjunction when it it is moving around data for wear-level purposes). An SSD itself internally only sees its memory blocks being occupied by what to it amounts as indeterminate data. An SSD is more sophisticated than a HD as it knows how many times a memory block has been re-written but this is not sufficient information for defragmentation.

While SSDs surely could use access patterns and wear-level analysis to try to guess what memory blocks belong to each other, such strategies would fail to cover all scenarios and fail when it comes to page files and other random writes (e.g. torrent downloads).

I just checked my system partition (running on an MTRON SSD) and the built-in defragmentation tool of XP says that I should defragment it (total fragmentation = 30%, file fragmentation = 59%).

I do admit that it is possible that I have misunderstood the complete abilities of current SSDs. It would be very nice if you could fire off an e-mail to one of your industry contacts to get a more detailed explanation.

 

I don't quite understand what you're saying when you talk about torrent downloads on the SSD, I'm getting a SSD drive some time soon and I'm an avid bittorrent user. Would getting a SSD drive make me have to change my download habits?

 

I was hypothesizing that if an SSD used access/write patterns and wear levels as predictors for internal defragmentation during wear-level operations, that it still would not work well for defragmentation. The reason would be the random write/read access patterns that occur in page files and torrent downloads, for example.

In regards to your question, I cannot tell if you need to change your download habits. I am not an avid bittorrent user but have on few occasions done heavy torrenting on my MTRON (which is SLC-based) without noticing any significant negative effects on system performance.

However, there have been reports that MLC-based SSDs (such as OCZ's Core series) have very poor random write performance. There are complaints saying that their Windows systems stall for seconds at a time due to poor handling of random writes. As torrent downloads involve random writes (which are the achilles heel of MLC-based SSDs) it could be a problem but I cannot tell for sure.

See this very interesting write-up on the issue. The technically oriented reader could actually do some calculations on how the benchmarked OCZ drive would perform during heavy torrenting, thanks to Steves calculations.

http://www.alternativerecursion.info/?p=106

 

Thanks for the link, I'll check it out, and thanks for all the information. I don't know if I'm going too off topic, but what's a MLC-based SSD? Is there another, more stable (as in good overall read/write times) kind of SSD out there as well? Uhh, maybe I should simply Google it, it's way too easy to just throw questions out.

I'm interested in SSD drives due to the near instant program startup, but if downloading with 1-2MB/s more or less constantly while performing other actions/gaming would be a problem for it, then I might reconsider.

Edit,
Seems I cant open the link you gave me, tried a few times, I'll try again later.

Edit #2,
The link works now.

 

Check out Les's thread on "Intro to SSDs" it explains a bit on MLC vs SLC and there are quite a few added links on more research - they helped me tremendously when considering buying my Sony Z with SSD

Linky

 

Hi,
if you are interested, here is a short discussion on defragmenting solid state drives
http://www.diskeeperblog.com/archives/2007/06/the_impact_of_f.html

That is incorrect.
This pdf may shed some light on the subject of interaction of filesystem and storage subsystem...
http://downloads.diskeeper.com/pdf/new-storage-technologies.pdf
(the first section)
and
http://files.diskeeper.com/pdf/HowFileFragmentationOccursonWindowsXP.pdf

Best regards
---------------------------------
Representative
Diskeeper Corporation
---------------------------------

 

Hi Diskkeeperguy,

Interesting read. The post by michael (with a small m) reflects my view (but I do not share his attitude ;-). For now, defragmenting flash drives without wear-leveling seems to be a good idea. The outcome of defragmenting wear-level enabled flash drives is unknown to the end user, as the external representation does not represent the internal data distribution of the SSD.

It would be very interesting if you could to perform the fragmentation/defragmentation test on an SSD with wear-leveling, using the internal Diskkeeper fragmentation tool and using the current Diskkeeper defragmentation software to defrag the drive and then publish the results.

In regards to the upcoming SSD defragmentation products, it would be interesting to see how they work. I assume that the firmware and wear-level algorithm of an SSD is the most valuable IP that an SSD-company has and that it would be hesitant to sharing it with others (even with Diskkeeper). They could however expand the ATA command set to enable software to gain direct access to actual data layout on the SSD. However, that would not be of use to existing SSDs.

When it comes to interesting developments, Managed Flash ( www.managedflash.com) are making some bold claims for their convert-random-writes-to-sequential-writes software. If their claims are correct, users of MLC-based SSDs might have a rescue in sight.

 

Hi JKetzetera,

Actually, Diskeeper is currently field testing SSD optimization with existing Diskeeper customers.

http://www.diskeeperblog.com/archives/2008/08/field_testers_n_1.html#comments

A white paper ought to be coming out later this year.


Best regards
------------------------------------
Representative
Diskeeper Corporation
--------------------------------------

 

Read your articles and kept seeing the reference to disks and hard drives, yet nothing about ssds. To state that I am incorrect without having some sort of supporting documentation is, well, premature. Its, of course, expected to see your salesmanship with respect to your product, however, I was dissappointed to later read that you havent yet done any conclusive testing with ssds.

The simple truth is, ssds do not slow with defragmentation and it does not occur in the same fashion as with a HD.

I say this only through my experience and that of several others I have spoken to here as well as in the consumer and enterprise manufacturing fields for ssds.

Quite frankly, wear levelling algorythms make products such as Diskeeper...antiquated.

Looking forward to your whitepaper when it comes out or any hands on experience that you may have to justify your thoughts.

 

Hi Les,
No offence intended, but i was addressing this statement:
"Much of the fragmentation of the hard drive is caused by the high speed of the spinning disk accompanied by the arm trying to pickup and place info."
File fragmentation is a function (er..'feature' ) of the file system, e.g. NTFS, FAT and not hard drive mechanics.

Best regards
---------------------------------
Representative
Diskeeper Corporation
---------------------------------

 

Also, here is an update that may be relevant to the thread. (Mods, please delete if you feel this post is inappropriate).

Partnership between Diskeeper and Apacer for SSD optimization.

http://www.diskeeperblog.com/archives/2008/09/its_flashtastic.html#comments

"As with other technologies, SSD's have had technical challenges to overcome, such as the limited number of erase-write cycles that undermine the SSD's durability. Apacer and Diskeeper specifically address this issue with key design elements employed in the development of the SSD+ Optimizer. By intelligently eliminating performance degrading free space fragmentation, HyperFast Flash technology actually reduces the aggregate erase-write cycles the SSD would otherwise incur during normal usage. In addition to performance benefits, SSD+ Optimizer offers increased SSD longevity."

Best regards
---------------------------------
Representative
Diskeeper Corporation
---------------------------------

 

Hi jketzetera

Another update on HyperFast that you might be interested in

http://downloads.diskeeper.com/pdf/HyperFast.pdf

Best regards
------------------------------------
Representative
Diskeeper Corporation
--------------------------------------