Sign in to follow this  
Roxor McOwnage

SSD's = Defrag is a thing of the past?

Recommended Posts

you are right. I sould have said: if no wear-leveling then the disk will die sooner because the same cells would be used over and over again.

Neverteless: some claim that MLC cells could only be used 3000 times instead of 10000 times.

Jeff

Share this post


Link to post
Share on other sites
IIRC, it is not possible to schedule defragmentation jobs with the built-in Disk Defragmenter utility, and I believe this behavior is intentional as an accomodation that MS reached with Executive Software.

Actually, we do it @ work with a scheduled task that runs at 5am daily, no extra utils needed. I believe we just use the command line to initiate it.

Before implementing that we got constant complaints of slowness from users that would go away once a defrag was run. After enough complaints from well, me and a few others who wanted to just buy Diskeeper or a similar product and deploy that, they came up with that and it's worked pretty well.

Share this post


Link to post
Share on other sites
you are right. I sould have said: if no wear-leveling then the disk will die sooner because the same cells would be used over and over again.

Neverteless: some claim that MLC cells could only be used 3000 times instead of 10000 times.

Jeff

The german C't magazin is testing MLC currently (and has been for some time). They stated they use a cheap 2GB USB stick with high transfer rate to archive lots of complete writes (the 2GB is to make sure it used recent MLC chips, and not some old crap).

They do a complete write, read, and md5 hash comparison. Currently, the el cheapo stick is at >20k cycles, and not a bit lost yet. Its getting a cycle every 6 minutes, so lets see when it will fail...

(And yeah, they showed that even cheap USB drives have wear leveling, as 10s of millions of writes to the "same" sector didnt do any damage)

Share this post


Link to post
Share on other sites

yes and my Cheetah's Sas drives should live for 114 years......but they will die long before that.

Here is interesting URLS's that not everything is as said.

http://www2.electronicproducts.com/MLC_and...p2008-html.aspx

i disagree however that heavy users use max 5 gb per day. When you go into hibernation or sleep you probable write some 2 gb and for waiking up the same.

So in my case this happens some 5 times a day, so i use per day some 20 gb without even working, right?

Jeff

Share this post


Link to post
Share on other sites
yes and my Cheetah's Sas drives should live for 114 years......but they will die long before that.

That has this to do with anything?

The same as you with your article in a german mag i guess....

i mean, even if any usb stick can be used forever, which it is not, there is no proof that flash SSD's will live as long as is claimed.

But if so, why only 2 years of warranty......

Jeff

Share this post


Link to post
Share on other sites
Some numbers for free space fragmentation and write performance from Diskeeper.

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

Looks like free space fragmentation can cause significant degradation of write performance in an SSD.

Thoughts?

Well Tenzen,

i am using SSD's right now for some 2-3 months without defragging and speed is always the same. You know this yourself if your computer slows down.

However, interesting to know this. I uninstalled Diskeeper.....

If ever, i will reinstall this of course.

Thanks

Jeff

Share this post


Link to post
Share on other sites
Some numbers for free space fragmentation and write performance from Diskeeper.

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

Looks like free space fragmentation can cause significant degradation of write performance in an SSD.

Thoughts?

Well Tenzen,

i am using SSD's right now for some 2-3 months without defragging and speed is always the same. You know this yourself if your computer slows down.

However, interesting to know this. I uninstalled Diskeeper.....

If ever, i will reinstall this of course.

Thanks

Jeff

Jeff, I am not really sure what 'Hyperfast' is. It does not appear to be a conventional Diskeeper defragger, but one that is meant specifically for SSDs. I am not even sure that it is available in a form other than as part of Apacer OEM hardware.

Unfortunately, I don't own any SSDs at the moment. They are beyond my justifiable price/GB ratio for now. I will snap one up for my laptop as soon as capacities hit > 160GB and prices drop to reasonable levels. ;) I am more interested in SSDs for the laptop (weight, power consumption, safety) rather than for the dekstop (where capacity is king)

Share this post


Link to post
Share on other sites

hi Tenzen,

i popped a question at the support of Diskeeper, since i am a member there. Waiting for their reply. It seems Raxco does the same, but i do not know how.

If you want an SSD for a laptop then you should wait until 2009 (my thoughts).

The OCZ Cores are having problems right now.

Still SSD's are way overpriced. Consider a Memoright 64 gb (faster then the 128 gb) for around $2000. They claim drives will stay good for 10 years. So this costs you $200 per year.

Other companies are saying their drives will last only 5 years.

It is not only the wear leveling that has to be considered (MTBF of 1.500.000 hours means nothing either).

Flash drives cannot hold the data forever. So once the drives are over 5 years best is to watch out for data failure.

Suppose you have a Raid5 with 6 drives: protection does not mean that much. If all of a sudden 2 drives fail......

Jeff

Share this post


Link to post
Share on other sites

Thanks for the info on the SSDs. Much appreciated.

I am going to adopt a wait and watch approach for now as far as SSDs are concerned.

$2000 for 64GB is simply out of the question for my personal usage. If I can persuade my boss to finance one for the work laptop, then price doesn't matter hehehe :lol:

The 10 year lifespan is not really useful for me, since the 64GB capacity will be quite obsolete in the near future itself. And who knows what interfaces will be around in 2018? :ph34r: . These low capacity drives will become legacy devices in a few short years IMHO.

hi Tenzen,

i popped a question at the support of Diskeeper, since i am a member there. Waiting for their reply.

Jeff

Do let us know if you get any more info from them.

Share this post


Link to post
Share on other sites

Okay, digging around the Diskeeper site a bit leads to a blog where you can ask questions directly to Diskeeper people.

There is also an article on the Hyperfast program. From what little I can make out, I think Hyperfast is already embedded into the Apacer SSDs and is not a retail product. So there is no need to 'install' it, let alone buy it separately or from the OEM. Interesting indeed.

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

Share this post


Link to post
Share on other sites

will do Tenzen,

but one thing is for sure. Hyperfast will defrag the entire partition, not only the free space. Right now, not such a good idea.

Some claim when writing to a SSD, 3 operations are involved (as explained here in another topic), that means that you do not have 10 000 REAL writes on your disk.....You do have the 10 000 writes, but you use 2-3 for one write. So that hurts.

I think my Cores will last 2-3 years, but by then the more lasting product will be available.

As for Diskeeper, i am still waiting for the support division to reply again.

Jeff

Share this post


Link to post
Share on other sites

ok, reply from Diskeeper Support:

"Thank you for the mail. Please note Diskeeper can defragment the free space on all the versions. It does it more effectively on the higher versions. Also along with this the data is also defragmented, which would reduce the erase write cycles of the SSD’s. We can enable the low free space defragment engine and run Diskeeper in short time frames to consolidate the free space fragments and reduce the impact on the SSD. I would recommend waiting for another month, when the new Diskeeper 2009 is released, it would have the functionality to detect the SSD state and consolidate the free space only. "

Jeff

Share this post


Link to post
Share on other sites
If the SSD is based on flash memory then defrag can help writing speed by keeping large contiguous blocks of open space available.

Flash is traditionally slow with writing because it is block-erased immediately before write operation. Having more contiguous open space available increases the odds that all the required space will already carry 1s or can be block-erased in one go, and that data bits will not need to be read then rewritten into a block.

This is completely incorrect.

More contiguous space as seen by the OS will have ZERO effect on contiguous space within the SSD unless it has almost no wear-leveling. Plus, defragging will force it to write more, don't do that.

If the OS sees a large contiguous block of free address space, that may be mapped to a ton of little bits here and there. There is no reason to defrag a SSD, EVER unless it doesn't have wear-leveling or bad block re-mapping, in which case you aren't running a consumer OS, you're dealing with firmware on a SSD or similar device.

Yeah, thats right, the firmware does the 'defrag' it needs to find or create empty blocks for writing on its own, in the background. A good one (like Intel's) is doing a lot of this sort of work behind the scenes.

There is only one, tiny benefit to defragging, the block level i/o requests can be for fewer large items rather than a large number of small requests.

But that is also pretty much pointless. I just measured 33,000 iops out of the intel drive for 4k random reads with 32 concurrent readers. And at 8k reads, I can push 220 MB / sec of pure random reads across a 32GB file on a system with 2GB RAM.

Don't defrag SSD's, internally, everything is fragmented and moving around in there and it won't have ANY effect on contemporary SSDs other than shortening their life span.

Share this post


Link to post
Share on other sites
ok, reply from Diskeeper Support:

"Thank you for the mail. Please note Diskeeper can defragment the free space on all the versions. It does it more effectively on the higher versions. Also along with this the data is also defragmented, which would reduce the erase write cycles of the SSD’s. We can enable the low free space defragment engine and run Diskeeper in short time frames to consolidate the free space fragments and reduce the impact on the SSD. I would recommend waiting for another month, when the new Diskeeper 2009 is released, it would have the functionality to detect the SSD state and consolidate the free space only. "

Jeff

Which of course, will do nothing for the newer drives, and only have a minor impact on drives with horrible wear-leveling algorithms (the JMicron based ones, perhaps).

Share this post


Link to post
Share on other sites
hi Tenzen,

i popped a question at the support of Diskeeper, since i am a member there. Waiting for their reply. It seems Raxco does the same, but i do not know how.

If you want an SSD for a laptop then you should wait until 2009 (my thoughts).

The OCZ Cores are having problems right now.

Still SSD's are way overpriced. Consider a Memoright 64 gb (faster then the 128 gb) for around $2000. They claim drives will stay good for 10 years. So this costs you $200 per year.

Other companies are saying their drives will last only 5 years.

It is not only the wear leveling that has to be considered (MTBF of 1.500.000 hours means nothing either).

Flash drives cannot hold the data forever. So once the drives are over 5 years best is to watch out for data failure.

Suppose you have a Raid5 with 6 drives: protection does not mean that much. If all of a sudden 2 drives fail......

Jeff

Even more ignorant misinformation. SSD's don't die like physical hard drives. With a hard drive, when it dies you basically lose EVERYTHING unless you go to a data recovery expert. With a SSD that dies due to writing too much, it simply enters read-only mode and disallows writes. Long before that happens SMART will have alerts and warnings that any decent OS or RAID controller will detect.

All drives can completely die if the connectors or chips on them break. But a SSD is SIGNIFICANTLY safer storage than a physical disk unless your data is precious enough to pay for a data recovery expert, which will know how to get stuff off a broken hard drive but getting stuff off a SSD requires more knowledge and expertise that just isn't around these days.

Hard drives often die within a few months to a year of use, and if they make it past that time will last for 10 ish years provided they aren't started and stopped too often or twisted axially violently or subjected to too many G-forces while they are turned on. Solid State drives, other than a few specific models that have had serious issues with quality, will simply work and last until THEY TELL YOU that they are dying, before they actually die.

So, if you have a raid 5 array and are stupid enough to ignore the SMART warnings, and not smart enough to access it in read-only mode to get data off, sure, you could have problems predictably in the future.

Share this post


Link to post
Share on other sites
Some numbers for free space fragmentation and write performance from Diskeeper.

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

Looks like free space fragmentation can cause significant degradation of write performance in an SSD.

Thoughts?

Only those with crappy controllers and wear-leveling algorithms. This trend won't last long and is an artifact of the current "'07" generation of SSD controllers.

Share this post


Link to post
Share on other sites
Thanks for the info on the SSDs. Much appreciated.

I am going to adopt a wait and watch approach for now as far as SSDs are concerned.

$2000 for 64GB is simply out of the question for my personal usage. If I can persuade my boss to finance one for the work laptop, then price doesn't matter hehehe :lol:

The 10 year lifespan is not really useful for me, since the 64GB capacity will be quite obsolete in the near future itself. And who knows what interfaces will be around in 2018? :ph34r: . These low capacity drives will become legacy devices in a few short years IMHO.

Are you ignoring the drive that is cheaper, and a lot faster, than that one while being larger?

Intels drive is going for ~$650 now, is 80GB, and is spec'd to last 5 years with 20GB of writes per day, but intel's own documentatin predicts that with that level of writing, the drive can last up to 15 years. And its MLC based . . .

I have tested it against a few other SSD's, all of them other than the Intel one and the most expensive 'server grade' SLC drives fail when there is read-write concurrency -- that is the reads degrade horribly while you write, and some writes can be really slow. But not the intel one -- I was doing a streaming write at 60MB/sec while completing 2000 random read iops, while the OCZ one slows to 10MB/sec writes while completing 300 reads/sec. The intel one will do ~ 7000 random writes/sec while many others can handle between 4 (Jmicron based) and 100(mid range server SLC) random write iops at most.

Later this year, to mid next year, other vendors will be putting out their next gen drives and many of these, will have also solved the random write and write concurrency performance issues of yesterday's SSD's.

Share this post


Link to post
Share on other sites
hi Tenzen,

i popped a question at the support of Diskeeper, since i am a member there. Waiting for their reply. It seems Raxco does the same, but i do not know how.

If you want an SSD for a laptop then you should wait until 2009 (my thoughts).

The OCZ Cores are having problems right now.

Still SSD's are way overpriced. Consider a Memoright 64 gb (faster then the 128 gb) for around $2000. They claim drives will stay good for 10 years. So this costs you $200 per year.

Other companies are saying their drives will last only 5 years.

It is not only the wear leveling that has to be considered (MTBF of 1.500.000 hours means nothing either).

Flash drives cannot hold the data forever. So once the drives are over 5 years best is to watch out for data failure.

Suppose you have a Raid5 with 6 drives: protection does not mean that much. If all of a sudden 2 drives fail......

Jeff

Even more ignorant misinformation. SSD's don't die like physical hard drives. With a hard drive, when it dies you basically lose EVERYTHING unless you go to a data recovery expert. With a SSD that dies due to writing too much, it simply enters read-only mode and disallows writes. Long before that happens SMART will have alerts and warnings that any decent OS or RAID controller will detect.

All drives can completely die if the connectors or chips on them break. But a SSD is SIGNIFICANTLY safer storage than a physical disk unless your data is precious enough to pay for a data recovery expert, which will know how to get stuff off a broken hard drive but getting stuff off a SSD requires more knowledge and expertise that just isn't around these days.

Hard drives often die within a few months to a year of use, and if they make it past that time will last for 10 ish years provided they aren't started and stopped too often or twisted axially violently or subjected to too many G-forces while they are turned on. Solid State drives, other than a few specific models that have had serious issues with quality, will simply work and last until THEY TELL YOU that they are dying, before they actually die.

So, if you have a raid 5 array and are stupid enough to ignore the SMART warnings, and not smart enough to access it in read-only mode to get data off, sure, you could have problems predictably in the future.

Thanks Scott, i did not know that. Now i feel more confortable using Raid0 with SSD's of course.

Neverteless, many companies claim that there drive will probably last for only 5 years, so in fact SSD's are very expensive if they do.

Jeff

Share this post


Link to post
Share on other sites

Yes Scott:

"Are you ignoring the drive that is cheaper, and a lot faster, than that one while being large"

if it was me i would be the Memoright 64 gb. So 2 of them instead of one 128 gb.

As for Intel, cheaper then the Memoright that's true.

My Cores costed me €308 a piece and i am happy with them.

Jeff

Share this post


Link to post
Share on other sites

he Scott,

"So, if you have a raid 5 array and are stupid enough to ignore the SMART warnings, and not smart enough to access it in read-only mode to get data off, sure, you could have problems predictably in the future."

I can understand that i make mistakes, which i will admit off course. We are all here to help each ohter, but your words above are not correct so it seems.

Here is the text on Anandtech:

IN THE NEAR FUTURE Intel will be releasing its own SSD tool that will let you query two SMART attributes on the drive: one telling you how close you are to the rated cycling limit, and one telling you when you've run out of reallocating blocks. The latter is the most important because Intel fully expects these drives to outlast their rated limits.

Meaning that the SMART feature is not on there drives YET.

Jeff

Share this post


Link to post
Share on other sites

I am curious whether the material used in the SSDs for bit storage will last intact (regardless of e/w cycles) without degradation at the microscopic level, for 10-15 years. I'll do some reading and get back on this issue later in the week. :unsure:

Share this post


Link to post
Share on other sites
Here is the text on Anandtech:

IN THE NEAR FUTURE Intel will be releasing its own SSD tool that will let you query two SMART attributes on the drive: one telling you how close you are to the rated cycling limit, and one telling you when you've run out of reallocating blocks. The latter is the most important because Intel fully expects these drives to outlast their rated limits.

Meaning that the SMART feature is not on there drives YET.

Hi Jeff,

I think what that quote means is that the SMART attributes already exist in the hardware, on the Intel drives at least, but that Intel feel that most users won't be able to query or interpret those attributes unless they (Intel) release a tool to do so. That could be because current versions of SMART reporting software don't report on this attribute, or because Intel is going to present it in a more understandable way. It may be that other SMART software will support this reporting capability in future versions.

I hope that other SSD vendors will also provide these SMART attributes from their drives, but given how (relatively) innovative Intel has been in its SSD products, it's conceivable that this feature is only available from Intel at present. Intel may well allow other vendors to use the same attributes in their drives, but it might be voluntary to adopt that (rather than being part of a new SATA standard for SSDs or something).

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
Sign in to follow this