Defragment SSD?

I have a 100 GB solid-state drive with not quite 60 GB used. The
Windows defrag tool (dfrgui.exe) says it is 7% fragmented. Since it is
solid-state, however, does defragmenting really mean anything?

The drive is Western Digital WD10EZEX-75ZF5A0.

--
David E. Ross

The Crimea is Putin's Sudetenland.
The Ukraine will be Putin's Czechoslovakia.
See http://www.rossde.com/editorials/edtl_PutinUkraine.html.

On Thu, 10 Jul 2014 11:42:03 -0700, David E. Ross wrote:

I have a 100 GB solid-state drive with not quite 60 GB used. The
Windows defrag tool (dfrgui.exe) says it is 7% fragmented. Since it is
solid-state, however, does defragmenting really mean anything?

The drive is Western Digital WD10EZEX-75ZF5A0.

The received wisdom is that defragging is not necessary for SSDs, and
moreover it can be harmful to them.

Think of finite allowed numbers of writes...

--
Gene E. Bloch (Stumbling Bloch)

On Thu, 10 Jul 2014 11:42:03 -0700, David E. Ross wrote:

I have a 100 GB solid-state drive with not quite 60 GB used. The
Windows defrag tool (dfrgui.exe) says it is 7% fragmented. Since it is
solid-state, however, does defragmenting really mean anything?

The drive is Western Digital WD10EZEX-75ZF5A0.

The main purpose of defraging is to minimize hard drive
head seek time. Since the SSD does not have heads, there
is no benefit.

--
Wildman GNU/Linux user #557453
The cow died so I don't need your bull!

On 10/07/2014 20:42, David E. Ross wrote:
I have a 100 GB solid-state drive with not quite 60 GB used. The
Windows defrag tool (dfrgui.exe) says it is 7% fragmented. Since it is
solid-state, however, does defragmenting really mean anything?

The drive is Western Digital WD10EZEX-75ZF5A0.


Yes, it does, like any other hard drive.
But I wouldn't do it, for two reasons:

1) Defragging causes a lot of r/w actions. On a SSD drive you wouldn't
like that to happen as it would wear out your drive. The number of write
actions is limited on a SSD.
It wouldn't make much difference anyway.

2) The MS defrag utility is not really capable.
May I cite Paul in his replies to my post "Does defragment service start
when idle or hybernation" on 08/05/2014:

"Be aware that the built-in defragmenter in the latest Windows is different.
It does *not* defragment files bigger than 50MB. It only handles
files smaller than that. This is a performance heuristic, where Microsoft
studied the performance impact, and decided what files made sense to
defragment and which ones to leave.

This causes a defragmenter run to complete faster. When
regularly scheduled and the task successfully gets a chance
to run, the residual fragmentation of small files is low,
and so each successive scheduled run won't have that
much to do.

If you are one of those people who "like the block colors"
and "want all blue blocks", you will not be happy about
what you see. For example, data partitions with movies on them,
the blocks could all be red (fully fragmented, no movie being in
one piece).

This is one of the reasons why the defragmenter actually
removed the "blocks" display. If you could see what
they are doing, you would freak out Microsoft's
decision is, "ignorance is bliss".

A third-party defragmenter will defrag all of them,
and make your partition "minty-fresh".

The free one I've used a couple times, was JKDefrag.

I tried the trial version of Raxco PerfectDisk, and
it seemed to work OK. But it costs money, and I don't
have money to waste on defragmenters. So JKDefrag would
have to do. The later version of JKDefrag is MyDefrag.
MyDefrag is still supported. "

With thanks to Paul

Fokke

David E. Ross said on 7/10/2014 2:42 PM:
I have a 100 GB solid-state drive with not quite 60 GB used. The
Windows defrag tool (dfrgui.exe) says it is 7% fragmented. Since it is
solid-state, however, does defragmenting really mean anything?

The drive is Western Digital WD10EZEX-75ZF5A0.

I just add the same for consensus of opinion: No moving heads and a
limited # of read writes. So don't waste the time.

Even though I did read an article (not on my finger tips at this moment)
that SSD's should last longer than the other computer hardware under
normal usage.

David E. Ross wrote:

I have a 100 GB solid-state drive with not quite 60 GB used. The
Windows defrag tool (dfrgui.exe) says it is 7% fragmented. Since it is
solid-state, however, does defragmenting really mean anything?

The drive is Western Digital WD10EZEX-75ZF5A0.

If access time to any part of memory is the same, what's the point of
defragmenting the memory? SSDs have graceful recovery for bad memory by
masking it to reserve memory. That masking takes more time and why an
SSD gets slower over time as it has to perform more and more masking due
to oxide stress on the transistors during writes. When there is no more
reserve space to mask bad memory, the device catastrophically fails. Do
your own research on "wear levelling". Defragging requires LOTS of
writing so a defrag on an SSD is not only superfluous but decreases the
lifespan of an SSD. See:

http://windows7themes.net/en-us/disa...windows-7-ssd/

While you should not run a defragmenter on the SSD, make sure that
background defragmentation is not enabled in the OS on the SSD and that
you're not using some defrag tool that is configured for background
defragging on the SSD. Besides scheduling defrags, some defrag tools
allow defragging while the computer is idle, like when the screen saver
activates (most I've seen do this by replacing the screen saver).

If you use a 3rd party defragger then you should not also use the one in
Windows. Different defraggers use different schemes to rearrange
clusters on the drive so you end up with them battling each other: one
undoes what the other does so the other one undoes the undoing when ran
again, and so on. Pick one defragger and stick with only its scheme for
defragging. Say you use the Auslogics' Defragger. Then you shouldn't
use the one in Windows, and that includes disabling the scheduled event
in Task Scheduler to run the Windows defragger (Task Scheduler Library
- Microsoft - Windows - Defrag, or run defrag from the Start menu
box). If you only use the Windows defragger then make sure the command
(action) for this event doesn't include your SSD drive; i.e., does not
include the drive letter you assigned to partition(s) on your SSD.

There is the Disk Defragmenter service (run services.msc and look at the
properties for this service). It is configured to Manual start mode
which means it is not started (loaded) until a process calls for it. I
wouldn't bother setting this service to Disabled since many defraggers
will use the standard system API defrag calls to do their defragging
(i.e., they have their own scheme for defrag but they still call the
defrag API provided by Windows).

http://computerstepbystep.com/disk_d...r_service.html

I don't know if disabling this service would interfere with defragging
by the Windows tool or a 3rd party tool. Since 3rd party tools use the
defrag API, I suspect disabling this service would have no effect on
those; however, since startup mode is Manual then it doesn't run until
something calls it.

http://www.howtogeek.com/howto/windo...windows-vista/
http://www.winhelponline.com/blog/de...ndows-7-vista/

Those mention how to disable the defrag service. When I try to run
defrag in the Start menu box, I get a warning that another tool has
taken over the Windows defrag settings (so I leave it alone since I use
a 3rd party defrag tool rather than the one in Windows). The dialog
lets you pick which volumes (drives) on which to defrag so you could
exclude the SSD. I suspect this is a wizard dialog to eliminate users
from having to delve into the Task Scheduler for the above mentioned
scheduled event. So you could use this dialog or Task Scheduler to
disable the background Windows defrag.

There is another Windows defrag performed while it is booting but its
effect is so tiny that it doesn't effect using a 3rd party defrag tool.
Plus, the 3rd party tools don't defrag the boot files (or have an option
to do so but on Windows startup) so there's no battling of the defrag
schemes (but it is superfluous on an SSD). See:

http://en.kioskea.net/faq/10353-disa...p-of-windows-7

http://www.faqforge.com/windows/defr...-on-windows-7/
mentions the undocumented /b parameter to defrag.exe but I don't see the
point of manually defragging the boot files when Windows will do this on
its startup, anyway. Yet, for an SSD, I don't see the point of trying
to defrag system files on Windows start because defragging of RAM is a
wasted effort.

Microsoft goes to great lengths to ensure users are getting their hard
disks defragged so even those that never manually run defrag.exe are
getting it ran for them, anyway. I don't that Microsoft was smart
enough in Windows 7 to disable its defragger on an SSD (and I have no
Windows 8 to look at right now to see if the BootOptimize on Windows
startup and the scheduled defrag event are disabled or exclude an SSD).

On 10/07/2014 20:14, Fokke Nauta wrote:
On 10/07/2014 20:42, David E. Ross wrote:
I have a 100 GB solid-state drive with not quite 60 GB used. The
Windows defrag tool (dfrgui.exe) says it is 7% fragmented. Since it is
solid-state, however, does defragmenting really mean anything?

The drive is Western Digital WD10EZEX-75ZF5A0.


Yes, it does, like any other hard drive.

But not anything useful.
Defragging an SSD will produce, at best, a tiny, probably barely
measurable, increase in speed.

Defragging a spinning disk speeds things up by reducing the amount of
wasted time as the heads move to find the next block of a file.
An SSD has no moving parts so defragging is pretty pointless.

--

Brian Gregory (in the UK).
To email me please remove all the letter vee from my email address.

David E. Ross wrote:
I have a 100 GB solid-state drive with not quite 60 GB used. The
Windows defrag tool (dfrgui.exe) says it is 7% fragmented. Since it is
solid-state, however, does defragmenting really mean anything?

The drive is Western Digital WD10EZEX-75ZF5A0.


The hard drive has a seek time of 15 milliseconds.

The SATA SSD has a seek time of 0.100 milliseconds (with
0.020 milliseconds of that being the time needed to
access a chunk of Flash and prepare it to be clocked out).
Flash is not as randomly accessible as DRAM is, and
it's kinda like getting an appointment at the
doctors office :-) Some delay is involved.

Some people are buying PCI Express Flash drives now,
and those will be getting closer to the bare minimum
0.020 millisecond number. Flash won't go faster than
that, in terms of random access, until someone takes
a hard look at "space" versus "performance" tradeoff.
And that's not going to happen any time soon.
"Space" is the only thing they care about.

In any case, it's a waste of time defragmenting the
SATA SSD drive. The SSD moves the heads 100x faster
than the mechanical hard drive. And when you defragmented
a mechanical drive, the performance improvement wasn't
that big a deal. If the drive is 100x faster moving the
heads, there just isn't a reason to bother.

*******

New owners of SSDs, owe it to themselves to spend around
a week reading articles on sites like Anandtech, to learn
what is good and what is bad for SSD drives. SSD drives
use one level of indirection, the internal storage
size does not map well to 4KB or 512 byte OS operations,
and so there can be internal fragmentation issues with
SSD drives. And that can cause a significant slowdown,
if the user "punishes" a drive with 4KB random write test.
That's what the review sites use, to better understand how
well drives recover from pathological write patterns. Intel
was one of the first SSD manufacturers to pay attention
to 4KB random write performance, while some other manufacturers
were content to have USB flash drive performance on theirs. It
pays to read up on articles like that, so you're getting
the very best from the drive.

Most end users, don't really have the punishing patterns
in their daily use. So they need not worry about this stuff.
But if a person starts asking "maintenance questions", then
yes, by all means, spend a week reading up on "best practices"
for SSDs. Just so you'll know what is bad for them, and
what activities are a "don't care".

I don't own an SSD, so I don't *have* to read the articles. Yet...

Paul

On 7/10/2014 11:56 AM, Wildman wrote:
On Thu, 10 Jul 2014 11:42:03 -0700, David E. Ross wrote:

I have a 100 GB solid-state drive with not quite 60 GB used. The
Windows defrag tool (dfrgui.exe) says it is 7% fragmented. Since it is
solid-state, however, does defragmenting really mean anything?

The drive is Western Digital WD10EZEX-75ZF5A0.

The main purpose of defraging is to minimize hard drive
head seek time. Since the SSD does not have heads, there
is no benefit.


While I agree with the statement about not degfragging...

I researched SSD's a while back and decided not to implement one.
There were numerous articles about SSD's being fast at first,
but slowing dramatically with use. Don't remember the term, but
there was a process to restore the speed. Don't think it was
defragmenting as we know it, but it did rearrange stuff in the flash
and required periodic fixup.
Had something to do with the fact that data is written in blocks.
When you have something in every block, your only option is to read
the block, change it, write it all back. Not the same process as
mechanical seeking, but has similar symptoms.

Maybe someone can enlighten us.

mike wrote:
On 7/10/2014 11:56 AM, Wildman wrote:
On Thu, 10 Jul 2014 11:42:03 -0700, David E. Ross wrote:

I have a 100 GB solid-state drive with not quite 60 GB used. The
Windows defrag tool (dfrgui.exe) says it is 7% fragmented. Since it is
solid-state, however, does defragmenting really mean anything?

The drive is Western Digital WD10EZEX-75ZF5A0.

The main purpose of defraging is to minimize hard drive
head seek time. Since the SSD does not have heads, there
is no benefit.


While I agree with the statement about not degfragging...

I researched SSD's a while back and decided not to implement one.
There were numerous articles about SSD's being fast at first,
but slowing dramatically with use. Don't remember the term, but
there was a process to restore the speed. Don't think it was
defragmenting as we know it, but it did rearrange stuff in the flash
and required periodic fixup.
Had something to do with the fact that data is written in blocks.
When you have something in every block, your only option is to read
the block, change it, write it all back. Not the same process as
mechanical seeking, but has similar symptoms.

Maybe someone can enlighten us.

TRIM is a feature which allows the OS to tell the
SSD drive, what sectors are no longer in usage. It
adds to the pool of free sectors that can be used
for data rearrangement.

The SSD drive itself, has one or two processors inside.
And firmware. The SSD works behind the scenes. In fact,
if you leave the computer running, the SSD can do write
operations to itself all night long. It you degraded the
SSD by doing a random 4KB write test ("hammered it"), a
good SSD will spend the whole night rearranging the blocks
to take up the least space. And leave as many whole free
blocks for tomorrow. By tomorrow, write performance will
be returned to the proper level again. Without such
behind-the-scenes maintenance, the write performance
might end up at 70% or 50% of the "good" value.

Why would it do that ? It consolidates small data objects,
in the larger flash structures. Flash storage structures
are inherently larger than the sizes that OSes like.
OSes may like 512 bytes or 4K bytes, and such things
are too small for flash. Maybe you use 4K bytes in an
flash area that is a megabyte in size. So the processors
inside the SSD drive, they rearrange the data, and pack
it better. The SSD has a level of indirection, a lookup
table, that maps external LBA, to internal flash location.
And so it can move things behind the scenes, and to the
external observer, they still can be read at the same LBA
as before.

Some of the Anandtech articles, go into proper technical
terminology for this stuff.

Paul

On Thu, 10 Jul 2014 17:54:50 -0700, mike wrote:

On 7/10/2014 11:56 AM, Wildman wrote:
On Thu, 10 Jul 2014 11:42:03 -0700, David E. Ross wrote:

I have a 100 GB solid-state drive with not quite 60 GB used. The
Windows defrag tool (dfrgui.exe) says it is 7% fragmented. Since it is
solid-state, however, does defragmenting really mean anything?

The drive is Western Digital WD10EZEX-75ZF5A0.

The main purpose of defraging is to minimize hard drive
head seek time. Since the SSD does not have heads, there
is no benefit.


While I agree with the statement about not degfragging...

I researched SSD's a while back and decided not to implement one.
There were numerous articles about SSD's being fast at first,
but slowing dramatically with use. Don't remember the term, but
there was a process to restore the speed. Don't think it was
defragmenting as we know it, but it did rearrange stuff in the flash
and required periodic fixup.
Had something to do with the fact that data is written in blocks.
When you have something in every block, your only option is to read
the block, change it, write it all back. Not the same process as
mechanical seeking, but has similar symptoms.

Maybe someone can enlighten us.

It has something to do with the fact that SSDs don't use the 1:1
mapping of system sectors vs. media sectors like hard drives.
SSDs use some sort of block mapping. I can't explain it because
I don't completely understand it myself. I don't use SSDs so I
haven't had the need to worry about it therefore I have done
little reading on the subject.

Also there is the issue of the limited number of read/write
cycles. Defragmenting would eat into that very quickly.

I think Paul had the best advice paraphrasing... research.

--
Wildman GNU/Linux user #557453
The cow died so I don't need your bull!

Wildman wrote:


I think Paul had the best advice paraphrasing... research.

My recommendation is, about a week of reading should cover it.
At one time, you could start with an ocztechnology.com article
from the forum, on tweaking an SSD. And end up knowing a bit.
But Anandtech and other places, delve into the disconnect
between traditional sector sizes, and the much larger
structures that flash are designed for. Keywords like
"SSD" and "TRIM" should get you pointed at some
good stuff.

It's always puzzled me, why someone in the industry didn't
rearrange the flash structures, to make the mapping a better
fit. So what if the density was cut to a half or only a
quarter. The density was recently "fixed", when
someone started stacking flash in 3D.

Also, research has shown, that "annealing" with heat,
will repair flash defects. The wearout phenomenon could be
almost completely removed, with annealing to fix
defects. But so far, there is no proposal on the
table, as to how this knowledge can be applied in
practice. A similar thing is known about LEDs,
namely that the intensity drops with time, and LEDs
can be restored by baking. But perhaps at a temperature
that would ruin the plastic packaging.

Paul

On 10/07/2014 23:20, Brian Gregory wrote:
On 10/07/2014 20:14, Fokke Nauta wrote:
On 10/07/2014 20:42, David E. Ross wrote:
I have a 100 GB solid-state drive with not quite 60 GB used. The
Windows defrag tool (dfrgui.exe) says it is 7% fragmented. Since it is
solid-state, however, does defragmenting really mean anything?

The drive is Western Digital WD10EZEX-75ZF5A0.


Yes, it does, like any other hard drive.

But not anything useful.
Defragging an SSD will produce, at best, a tiny, probably barely
measurable, increase in speed.

Fully agree. I guess you wouldn't tell the difference.


Defragging a spinning disk speeds things up by reducing the amount of
wasted time as the heads move to find the next block of a file.
An SSD has no moving parts so defragging is pretty pointless.

On 10/07/2014 2:42 PM, David E. Ross wrote:
I have a 100 GB solid-state drive with not quite 60 GB used. The
Windows defrag tool (dfrgui.exe) says it is 7% fragmented. Since it is
solid-state, however, does defragmenting really mean anything?

Nope, defragging means nothing to an SSD. Don't even look at the defrag
report.

Yousuf Khan

"Paul" schreef in bericht
...
mike wrote:
On 7/10/2014 11:56 AM, Wildman wrote:
On Thu, 10 Jul 2014 11:42:03 -0700, David E. Ross wrote:

I have a 100 GB solid-state drive with not quite 60 GB used. The
Windows defrag tool (dfrgui.exe) says it is 7% fragmented. Since it is
solid-state, however, does defragmenting really mean anything?

The drive is Western Digital WD10EZEX-75ZF5A0.

The main purpose of defraging is to minimize hard drive
head seek time. Since the SSD does not have heads, there
is no benefit.


While I agree with the statement about not degfragging...

I researched SSD's a while back and decided not to implement one.
There were numerous articles about SSD's being fast at first,
but slowing dramatically with use. Don't remember the term, but
there was a process to restore the speed. Don't think it was
defragmenting as we know it, but it did rearrange stuff in the flash
and required periodic fixup.
Had something to do with the fact that data is written in blocks.
When you have something in every block, your only option is to read
the block, change it, write it all back. Not the same process as
mechanical seeking, but has similar symptoms.

Maybe someone can enlighten us.

TRIM is a feature which allows the OS to tell the
SSD drive, what sectors are no longer in usage. It
adds to the pool of free sectors that can be used
for data rearrangement.

The SSD drive itself, has one or two processors inside.
And firmware. The SSD works behind the scenes. In fact,
if you leave the computer running, the SSD can do write
operations to itself all night long. It you degraded the
SSD by doing a random 4KB write test ("hammered it"), a
good SSD will spend the whole night rearranging the blocks
to take up the least space. And leave as many whole free
blocks for tomorrow. By tomorrow, write performance will
be returned to the proper level again. Without such
behind-the-scenes maintenance, the write performance
might end up at 70% or 50% of the "good" value.

Why would it do that ? It consolidates small data objects,
in the larger flash structures. Flash storage structures
are inherently larger than the sizes that OSes like.
OSes may like 512 bytes or 4K bytes, and such things
are too small for flash. Maybe you use 4K bytes in an
flash area that is a megabyte in size. So the processors
inside the SSD drive, they rearrange the data, and pack
it better. The SSD has a level of indirection, a lookup
table, that maps external LBA, to internal flash location.
And so it can move things behind the scenes, and to the
external observer, they still can be read at the same LBA
as before.

Some of the Anandtech articles, go into proper technical
terminology for this stuff.




SSD may be faster, but I believe that life span of an SSD is stil less than
that of a conventional hard disk? That's the main reason for me not to touch
SSD for the time being...



--


|\ /|
| \/ |@rk
\../
\/os