Status
Not open for further replies.

OldRick

Distinguished
Jul 11, 2008
3
0
18,510
This may be a silly question, but I thought I'd check.

I'm now running off a new Mtron SSD. It's very fast, and shows about 86MB/s read speed on my system. HDtach shows random access time of .1ms.

The question whether there is any reason whatsoever to defragment such a drive?

Seems like if the random access time from any one block to the any other block is always .1ms, then it wouldn't matter in the slightest where the blocks are located, except for the overhead of managing fragments and tracking them in the bitmap.

So perhaps the real question is whether the overhead of managing fragmented files on a storage volume is noticeably greater than managing non-fragmented files?

Can anyone supply any insight on this?
 

OldRick

Distinguished
Jul 11, 2008
3
0
18,510
You're probably right, but it's hard to give up something I've been so compulsive about for the last 25 years...
 
Yes, but in all likelihood, something that the operating system sees as "sequential" will be mapped to relatively random locations in the FLASH anyways, due to wear leveling techniques used in current SSDs. So, you'll be doing a whole bunch of reading and writing to move stuff from one random order to another random order, causing wear to the flash in the process.

(I religiously defrag too, but I don't have the money for a good SSD yet)
 

orangegator

Distinguished
Mar 30, 2007
1,163
0
19,310


Yeah, this also what I've read. You want the data spread throughout the drive, not just part of it where that section wears out quicker due to constant reads and writes. Since there aren't any spinning platters in a ssd, there is no speed advantages to defragging.
 

chuckt

Distinguished
Mar 7, 2007
72
0
18,630
• Flash Cell Endurance: For Multi-Level Cell (MLC) Flash, up to 10,000 write cycles
per physical sector. For Single-Level Cell (SLC) Flash, up to 100,000 write cycles per
physical sector.
According to Toshiba, the inventor of Flash memory: “the 10,000 cycles of MLC NAND is more than sufficient for a wide range of consumer applications, from storing documents to digital photos. For example, if a 256-MB MLC NAND Flash-based card can typically store 250 pictures from a 4-megapixel camera (a conservative estimate), its 10,000 write/erase cycles, combined with wear-leveling algorithms in the controller, will enable the user to store and/or view approximately 2.5 million pictures within the expected useful life of
the card.”1
For USB Flash drives, Toshiba calculated that a 10,000 write cycle endurance would enable customers to “completely write and erase the entire contents once per day for 27 years, well beyond the life of the hardware.”

http://www.kingston.com/products/pdf_files/FlashMemGuide.pdf
 

chuckt

Distinguished
Mar 7, 2007
72
0
18,630
I would suggest getting some real figures from the manufacturer and test all claims and assumptions including having your own backup plan but I'm guessing that you have a greater chance of the SSD burning out before you write to it too much which is essentially an oxi-moron argument.

As a sanity check - I found some data from Mtron (one of the few SSD oems who do quote endurance in a way that non specialists can understand). In the data sheet for their 32G product - which incidentally has 5 million cycles write endurance - they quote the write endurance for the disk as "greater than 85 years assuming 100G / day erase/write cycles" - which involves overwriting the disk 3 times a day.

http://www.storagesearch.com/ssdmyths-endurance.html
 

chookman

Distinguished
Mar 23, 2007
3,319
0
20,790
The 2 main reasons for defragging your disk is to limit the amount of times the drive heads have to move to a different section of the disk to read the data (limit the use of seek times on the disk). The second is to place the majority of your data on the outer edge of the disk where your read and write speeds are faster on a mechanical drive do to simple mechanical physics.

SSD's elimate both of these, one their seek times are next to nil. So moving to a different section of the disk causes little to no delay in data read/writes so you dont need to have a file placed sequentially on the disk. And their is no drop in performance over the entire size of the disk because there is no mechanical limiting factor continually writing everything to a particular section on an SSD will actually decrease the life expectancy on the drive as there are only so many writes per cell before failure. As listed by chuckt above.
 

moonwave

Distinguished
Aug 2, 2009
20
0
18,510
Being that the life span of an ssd drive is supposed to be more than a hard drive, I don't see the harm in defragmenting them if not too frequent. That said, I've seen horrible fragmentation in just a few months. So I couldn't imagine using a drive for 5 years without ever fragmenting it. I also frequently restore HD's from backups which is an entire HD copy. So either way drives get used and if SSD's where that flimsy they'd be worthless. But since the life of an SSD is greater than a HD, again it shouldn't be a problem.

I recently had to defrag a computer for someone running Wow. The result of playing the game for over a year with no defrag. It was stuttering, crashing and had all sorts of problems. The page file was in a bazillion pieces. But one defrag made the game smooth again. A lot of people only pay attention to the concept that the ssd has zero seek time. So think therefore defragging shouldn't help, however I still believe there are benefits. Each time data is not contiguously stored, the OS must contend with pointers to data in other locations. It's not much overhead, but it's not zero. Also having every file fragmented into a billion pieces could eventually lead to the disk structure crashing. Hopefully not, but who's to say it's perfect. :na: :lol:

However someone who loves to defrag every week would be over doing it with SSD. A smarter move would be to defrag once or twice a year. That way the amount of usage isn't very much. But I certainly wouldn't let my partition go 5 years without ever defragging. It would be like swiss cheese, lol. I've done testing on RAM drives in the past to see the effects of adding 5,000, 10,000, 20,000 small files and even though there was no physical disk. Windows got slower and slower in just trying to index them. So it's really no different than any other piece of software you have. There is no physical arm to move on the disc, but software takes time to execute. Simplify what the software needs to do and it will run faster. I need to do more testing, but I suspect that a horribly fragmented SSD will perform more slowly. And if not, it could eventually lead to other problems.

But let's suppose it really made no difference. Then by defragging, the files are still stored in a more logical fashion. And if that didn't matter? Then it still won't matter if the user defrags because he only uses a small portion of the SSD's life that is already longer than that of a HD. So, my recommendation is to do infrequent defrags on SSD's. But they are quite a bit more robust in terms of MTBF as someone here already suggested. My guess is 2 or 3 years out, people will defrag their SSD's anyway, and it won't matter. And sure you can extend the life of a SSD by not defragging, but it hardly matters if it's infreuqently. A Hd gets wear too either way. Point is I figure people are overly protective about SSD's because they are a new and expensive toy. But that will all change soon. So a little wear isn't a big deal. Today's $330 toy is worth $33 or less in 5 years anyway. Also check the recent prices. 64GB for only $129? 32GB for $80?? Not bad!!! So just enjoy the speed :D
 

sub mesa

Distinguished
Since you opted to resurrect this thread, let me add:

Modern SSDs like Intel X25-M use a trick to get high random write speeds. When defragmenting, it will overwrite existing data. For example, it will put the most used applications at the beginning of the disk, as that's where HDDs are faster. But in reality, the Intel SSD has remapped the locations to free flash cells in order to get higher performance.

So what windows thinks is the beginning of the disk, may be somewhere in the middle. The SSD is silently moving data to other flash cells without Windows or anything else ever knowing. Now here comes the trick: it has to remember where all these 'mappings' correspond to. For every I/O this 'list' needs to be referenced, to see where the data really is being stored. But with defragmenting this list can grow enormously, and this starts to hamper performance. To remedy, all sectors need to be zero-written again, or using a special utility to wipe the special flash cells on Intel SSDs.

So if you defragment, this 'list' is getting filled up and can slowdown performance significantly. So defragmenting can actually be very harmful to performance. So there should be no reason to do it. :)
 

sidewinderdt

Distinguished
Jan 16, 2009
35
0
18,540
The drive's controller/firmware decides which cells/blocks to use for the storage of data; the file system works only at the logical disk level, and has no say in the actual physical placement of the data on the drive, be it a SSD or a HDD. So, filesystem fragmentation is different from the 'internal' fragmentation that you have described. One has no direct connection to the other.:)

From what I've read, defragging an SSD is not very useful, but occasional free space consolidation is, in reducing the propensity for random writes and improving write performance after the drive has seen some use. Hyperfast which is the SSD optimizer add-on for Diskeeper 2009, is said to do a good job.
http://www.diskeeper.com/blog/post/2008/12/02/HyperFast-is-also-here!.aspx
 

sub mesa

Distinguished
The filesystem fragmentation will have little effect to the actual speed or IOps rate, but having the write remapping list very full will cause significant slowdown across the board.

From what i understand, the TRIM function in Windows 7 will make that Disckeeper software obsolete, since free-space reclamation is done automatically when you delete a file. Using free privacy tools like Eraser might also allow to zero-write deleted items, instead of leaving random garbage behind.
 
G

Guest

Guest
I would encourage everyone interested to read this article about SSD's
http://www.anandtech.com/storage/showdoc.aspx?i=3531&p=1
It is long, and gets very technical, but it definitely reads as a well researched and very accurate article.
The overall is that defrag can be detrimental, but not because it causes wear as much as it does not zero out a cell. You would need to run some kind of special tool (not sure that it exists other than a special erase all disk erase tools) in order to improve performance (as I think some have already mentioned).

Instead, based on the design of SSD's, I would be interested in a tool that would allow a file to be "striped" across the internal controllers so that it could read files across multiple paths to different memory at the same time. That seems to me like a real way to improve SSD performance. minimize the delay of the controller only being able to charge 1 cell at a time and look for information by using however many controllers are in the system at the same time. I am not sure whether this would be feasible, but it certainly sounds good in my head...
 

vh1atomicpunk

Distinguished
Apr 24, 2008
97
0
18,640

windstrings

Distinguished
Mar 15, 2010
1
0
18,510
We aren't dealing with a physical disk and a physical arm that has to reach over and access the physical location. Its all non moving solid state.

Defragging solid state is getting all worked up with no place to go... the flash thumb drives would wear out if you defragged and totally formatting and recopying the info back over was best for flash thumb drives.

But SSD drives have top notch technology that takes normal use into account. If you start defragging it.. thats not normal use anymore and you may in fact be frustrating the indexing system and slowing it down as well as putting unnecessary wear on it.

How fast is fast?... when your accessing that fast.... even if you double it, or half it, it will only be notiecable hardware utilities...... Worst case scenario and you still kill a regular hardrive for speed.
 

philswift

Distinguished
Jun 1, 2010
1
0
18,510


No one knows for sure until some scientific tests are done. SSD drives are upto 256GB now so it seems that they will take over HDD. Someone needs to have 3 identical PC's with identical SSD's under similar user loads, run one for 12 months having no disk optimisation (turn off any defrag), run one for 12 months having OS defrag turned on and run one using MyDefrag (industry standard HDD optimisation (not just defrag)) and see what happens. Even when you build up a new PC it can be heavily defragged. Any user perceived performance improvement is worth losing some service life, especially as it seems that SSD life is longer than HDD under normal Enterprise user workloads.

 

groucho7

Distinguished
Oct 2, 2010
5
0
18,510
Intel clearly recommends NOT to defragment their SSD. You have had many answers telling you in great detail why this is so. If you are so compulsive about defragging that you can't help doing it to your SSD you would do better to see a shrink than to defrag your precious SSD and ruin it fast :)
 

groucho7

Distinguished
Oct 2, 2010
5
0
18,510



Intel says NO.
 

graham006

Distinguished
May 3, 2009
12
10
18,515


Take up smoking. It will be better for your hard drive.
 
Status
Not open for further replies.