Don't Defrag Your SSD

equalsql

It appears there is almost zero performance gain, and it just serves to eat into the precious finite write cycles that an SSD has.
(i.e. SSDs tend to just 'vanish' whilst traditional hard drives give you some indication that things are about to go wrong)

http://www.pcworld.com/article/2047513/fragging-wonderful-the-truth-about-defragging-your-ssd.html?page=1

Interesting stuff.

digitalscream

Yeah, that's been common knowledge since SSDs were released.

The only reason that fragmentation causes performance issues on traditional spinning-rust drives is because of rotational latency and by extension how long it takes to find data on the drive. That's basically irrelevant in the solid state world.

xmrchi

Plus it can damage the drive,

Myranda

digitalscream said:

Yeah, that's been common knowledge since SSDs were released.

The only reason that fragmentation causes performance issues on traditional spinning-rust drives is because of rotational latency and by extension how long it takes to find data on the drive. That's basically irrelevant in the solid state world.

This!

Twice.

I think it would be an idea if disk defrag utilities started to post the fact that SSDs don't need defragging in big letters on loading

equalsql

Myranda said:

digitalscream said:

Yeah, that's been common knowledge since SSDs were released.

The only reason that fragmentation causes performance issues on traditional spinning-rust drives is because of rotational latency and by extension how long it takes to find data on the drive. That's basically irrelevant in the solid state world.

This!

Twice.

I think it would be an idea if disk defrag utilities started to post the fact that SSDs don't need defragging in big letters on loading

I agree, I've never actually owned a 'true' SSD so I was not aware of this issue.

digitalscream

equalsql said:

I agree, I've never actually owned a 'true' SSD so I was not aware of this issue.

Thing is, in the majority of situations it's actually the lack of rotational latency that really speeds up machines using SSDs, not the silly transfer rate numbers. Most hard drive reads/writes are relatively small files, so it's the seek times (ie looking for a bit of data on the disk) that punish you when you're using a traditional drive. That's why defragging helps in that situation - it just puts all the bits of all the files in a row so the drive heads don't need to be repositioned to get the next bit of the file.

Good defragging utilities also tend to put the most frequently-used files together in the fastest (outer-most) part of the drive. This tends to be the page file (virtual memory) and operating system files.

Fazer

i knew about not needing to defrag ssds, but what about the finite write cycles? is it any more finite than a regular spinning harddrive?

digitalscream

Fazer said:

i knew about not needing to defrag ssds, but what about the finite write cycles? is it any more finite than a regular spinning harddrive?

Kinda...it's quite hard to get meaningful comparisons compared to regular hard drives - for example, a modern SSD will last around 5 years rewriting every single block on the drive in a random pattern every half hour. Even with the absolute fastest regular drives out there, that's basically impossible to manage.

No, they're not going to last forever, and for data I feel is really important I'll still store it on a traditional drive. Then again...there isn't much digital storage that would last forever. However, in five years' time you'll most likely have ditched that drive for something else.

mudslide73

They're brill. I've been replacing laptop hdds with ssds for a couple of years. Very good at extending the perceived life of a machine by increasing the performance. Never thought about defragging at all though.

digitalscream

mudslide73 said:

They're brill. I've been replacing laptop hdds with ssds for a couple of years. Very good at extending the perceived life of a machine by increasing the performance. Never thought about defragging at all though.

Not only do they extend the perceived life, they extend the battery life too (up to an extra hour for a laptop that can do 4 hours away from the mains, in my experience).

equalsql

Fazer said:

i knew about not needing to defrag ssds, but what about the finite write cycles? is it any more finite than a regular spinning harddrive?

There's a good article on Tom's Hardware about the robustness of SSDs here

My experience of failures on them is that you get no warning, unlike hard disks that start to throw sector errors or start making mechanical noise. Other's of course may have experienced things differently.

digitalscream

equalsql said:

Fazer said:

i knew about not needing to defrag ssds, but what about the finite write cycles? is it any more finite than a regular spinning harddrive?

There's a good article on Tom's Hardware about the robustness of SSDs here

My experience of failures on them is that you get no warning, unlike hard disks that start to throw sector errors or start making mechanical noise. Other's of course may have experienced things differently.

That article is a bit dated - given that SSDs had only been around in the mainstream for about 2 years at the time. Most of the issues - including the terrifying firmware bugs which destroyed a lot of drives back then - have been fixed. Write levelling technology has been improved, and operating system support for extending the life of SSDs is now commonplace.

mudslide73

I've had one doa and one fail in the wild out of about 70 which is pretty high. The faulty one just bricked without warning - all Sandisks.

digitalscream

mudslide73 said:

I've had one doa and one fail in the wild out of about 70 which is pretty high. The faulty one just bricked without warning - all Sandisks.

Really? In my experience, that's pretty low for traditional drives. From personal experience, I'd expect at least two or three times that.

Fuzzdog

digitalscream said:

mudslide73 said:

I've had one doa and one fail in the wild out of about 70 which is pretty high. The faulty one just bricked without warning - all Sandisks.

Really? In my experience, that's pretty low for traditional drives. From personal experience, I'd expect at least two or three times that.

Agreed - that's a hell of a lot lower failure rate than I've had with spinning rust drives, and that's speaking from fairly low-stress installations.

Love my SSD, it was one of the best upgrades I ever made.

xmrchi

I have had a OCZ ssd fail after 1.5 years, as someone said above, you don't get any clue it's going to die unlike a platter drive.

Axe_meister

I thought that running defrag on an SSD just initiates the trim command.
Certainly that is what happens on. OSX and win 8

mudslide73

I think that's what I mean.. they've gone pretty quickly. I must have been lucky with mechanical drives because I've not had many go bad until they're old (5 years +).

I was a field engineer when Fujitsu 3.5 drives failed en masse in 2003 ish. I must have replaced 200 just in the Midlands.

digitalscream

mudslide73 said:

I was a field engineer when Fufitsu 3.5 drives failed en masse in 2003 ish. I must have replaced 200 just in the Midlands.

Heh...I worked tech at a big insurance company in the late 90s to very early 2000s. I vividly remember doing a complete desktop refresh on a couple of thousand PCs shortly before the IBM DeskStar apocalypse arrived. Guess which drives we had in all our desktops.

Axe_meister

You should see the number of dead drives in your average data centre.
We have a daily runner in each of our DC, just replacing drives. Thank god for RAID and hot swap

Jez6345789

Hi 

There are a number of good tools out there to help get reliability stats from your SSD some included, I use Drive DX. I have lost hard drives in 2 years so that is my benchmark to swap the drive or move it to a less critical position.

My old Mac OCZ fitted for 2 years and it showed no potential failures after daily 16 hr day use sometime.