Is ext4 bad for lifetime of SSD?

aardwolf · 2014-09-18 11:31:00

Hi,

A post here says ext4 will half lifetime of an SSD: http://askubuntu.com/questions/75061/sh … for-my-ssd
The Arch wiki seems to be in favor of ext4 for an SSD however: https://wiki.archlinux.org/index.php/So … Filesystem

Since there are conflicting opinions at different times, is there anyone with knowledge on the subject who knows whether ext4 is bad for SSD lifetime?

Thanks!

TheSaint · 2014-09-18 11:36:19

Why don't we make a poll to whom has a SSD and for how long.

Trilby · 2014-09-18 12:01:55

So a user who posts anonymously on askubuntoo.com refers to "something they read once on the internet" while on a cigar shop website ... yeah, that sounds authoritative. I'd prefer to ask my magic 8 ball.

edit: I have my full system on an SSD using ext4 for / and /home and ext3 for /boot. This is, of course, not evidence of anything either - just contributing to the 'poll' side of this.

Last edited by Trilby (2014-09-18 12:03:36)

WorMzy · 2014-09-18 12:46:03

All filesystems are bad for SSDs. The best way to maximize the life of your SSD (or HDD) is to leave it in the antistatic bag.

I've had my SSDs for a year or two now, with btrfs and ext4 and even ntfs and fat filesystems on them. I've not had any problems.

xedchou · 2014-09-18 12:53:36

aardwolf wrote:

Hi,
A post here says ext4 will half lifetime of an SSD: http://askubuntu.com/questions/75061/sh … for-my-ssd
The Arch wiki seems to be in favor of ext4 for an SSD however: https://wiki.archlinux.org/index.php/So … Filesystem
Since there are conflicting opinions at different times, is there anyone with knowledge on the subject who knows whether ext4 is bad for SSD lifetime?
Thanks!

According to my memory, nobody did the test. So just use it.

alphaniner · 2014-09-18 13:16:47

All of the recent articles I've read indicate that concerns about SSD lifetime are overblown, and have been for a while now. Basically, quality SSDs will become obsolete before they fail (eg. they can handle GBs of writes per day, every day, for several years).

Just steer clear of generic SSDs from Chinese sellers on eBay and get on with your life.

fsckd · 2014-09-18 14:06:37

aardwolf wrote:

Hi,
A post here says ext4 will half lifetime of an SSD: http://askubuntu.com/questions/75061/sh … for-my-ssd
The Arch wiki seems to be in favor of ext4 for an SSD however: https://wiki.archlinux.org/index.php/So … Filesystem
Since there are conflicting opinions at different times, is there anyone with knowledge on the subject who knows whether ext4 is bad for SSD lifetime?
Thanks!

That guy's answer is pretty bad. He hasn't given any link to an article when he was asked. There's nothing substantial other than a list of claims with nothing to back it up. How do we know he wasn't using a crappy product from a shoddy manufacturor?

Another answer there suggests using ext3 instead. That's seriously messed up advice.

My advice to you is to do a little more Googling before accepting some random stranger's "anecdotal evidence" from years ago.

R00KIE · 2014-09-18 15:48:14

If the SSD's firmware is any good and does static wear leveling (in addition to dynamic wear leveling) there should be no problems. There are tests where some SSDs were good for many terabytes[1], so in normal use they should last many years.

I'd be more concerned about data retention, however I have never looked into that and I don't know if anyone has yet. By this I mean, if you write once and very seldom to the ssd, how long will the data remain readable.

[1] http://techreport.com/review/26523/the- … a-petabyte

smlb · 2014-09-18 19:09:21

Full ext4 system on SSD, works like a charme.

nstgc · 2014-09-18 20:01:58

From everything I've seen, I'd say that most file systems that aren't too ancient will be equally harsh on an SSD. The newer ones, such as ZFS, btrfs as well as the one that Microsoft is developing, are easier on them. Still, by the time a modern SSD is worn down, you'll likely want to replace it for other reasons anyway, so I wouldn't worry about which FS you use to format your SSD with so long as it supports TRIM.

defears · 2014-09-18 21:30:19

I've had my SSD for about 3 to 4 years now and still working fine. Only ever used ext4 with one partition because it's only 60 Gigs.

emeres · 2014-09-18 22:36:00

Context:
http://www.electronicproducts.com/Compu … tions.aspx
http://www.ni.com/white-paper/10126/en/

Papers:
https://www.usenix.org/legacy/event/use … .bak1.html

http://ieeexplore.ieee.org/xpl/articleD … er=6176473
http://ieeexplore.ieee.org/xpl/articleD … er=6417326
http://ieeexplore.ieee.org/xpl/articleD … er=6560757

@R00KIE Chapter 3.4 might be of interest to you.

@OP Should you have no access to those papers, ask someone who has. Depending on subjective observations is highly questionable. R00KIE provided something tangible, but those reviewers should be beaten with a bamboo stick for those "graphs".

Side note: Some information might be more than just a little outdated by now and I apologize for any ignorance displayed concerning the topic.

scryan · 2014-09-19 02:23:28

I read scientifically on the internet that now-a-days SSD lifespans have gotten longer then HDDs.
Sounds good, so I am going with that. Been working out so far...

ewaller · 2014-09-19 02:59:11

I have not been involved with any designs using SSD for more than four years, so I have nothing definitive to say anymore as to the current state-of-the-art. But, I can say that not all SSDs are created equal, so ensure that you are comparing Apples to Apples. One of the biggies that might bite you on the butt is multi level cell memory versus single level cell memory

In multi level Cell memory, you store more than one bit per transistor. Using four or eight analog levels per cell, they can code for 2 or 3 bits. The error rates of these memories are higher, and have lower write endurance. In single level cell memory, you store one bit per transistor (as does DRAM, EPROM, EEPROM) with one threshold, not 4 or 8. This leaves a lot more room for deterioration, but provides a much lower storage density.

Last edited by ewaller (2014-09-19 03:00:15)

ugjka · 2014-09-19 09:09:15

Always remember keep your ssd's firmware up to date. Which would be simple if ssd's manufacturers provided linux binaries for updating the firmare. I had to pirate live windows 7 iso just to do that. I hated doing that but I had no choice

teateawhy · 2014-09-19 09:59:34

ugjka wrote:

Always remember keep your ssd's firmware up to date. Which would be simple if ssd's manufacturers provided linux binaries for updating the firmare. I had to pirate live windows 7 iso just to do that. I hated doing that but I had no choice

You can download a dumbed down version of windows called Windows PE for free, which allows you to do firmware updates.
http://en.wikipedia.org/wiki/Windows_Pr … nvironment
Of course i agree that providing linux binaries, or source code, would be better.

nstgc · 2014-09-19 16:42:09

teateawhy wrote:

ugjka wrote:
Always remember keep your ssd's firmware up to date. Which would be simple if ssd's manufacturers provided linux binaries for updating the firmare. I had to pirate live windows 7 iso just to do that. I hated doing that but I had no choice
You can download a dumbed down version of windows called Windows PE for free, which allows you to do firmware updates.
http://en.wikipedia.org/wiki/Windows_Pr … nvironment
Of course i agree that providing linux binaries, or source code, would be better.

Thanks! I've never heard of that.

There is the Arch wiki article: https://wiki.archlinux.org/index.php/Windows_PE
And from Microsoft: http://technet.microsoft.com/en-us/libr … s.10).aspx

ugjka · 2014-09-19 16:56:38

nstgc wrote:

teateawhy wrote:
ugjka wrote:
Always remember keep your ssd's firmware up to date. Which would be simple if ssd's manufacturers provided linux binaries for updating the firmare. I had to pirate live windows 7 iso just to do that. I hated doing that but I had no choice
You can download a dumbed down version of windows called Windows PE for free, which allows you to do firmware updates.
http://en.wikipedia.org/wiki/Windows_Pr … nvironment
Of course i agree that providing linux binaries, or source code, would be better.
Thanks! I've never heard of that.
There is the Arch wiki article: https://wiki.archlinux.org/index.php/Windows_PE
And from Microsoft: http://technet.microsoft.com/en-us/libr … s.10).aspx

So there was a choice afterall. If only I asked in the forums... Duh.

R00KIE · 2014-09-19 19:35:30

@emeres
Why do you say the graphs are not good?
The paper you link to looks interesting, however it still doesn't answer my doubts about data retention. When I have time I might look into this, I'm sure I'll learn a few more things.

@ewaller
What you say is true but it seems manufacturers are starting to come up with clever ways to optimize/squeeze every last drop of useful life out of TLC flash [1]. For consumer grade drives I guess things are reliable enough, for enterprise I guess no one wants anything else than SLC.

[1] http://anandtech.com/show/8520/sandisk- … ssd-review

emeres · 2014-09-19 21:01:52

@R00KIE Those images qualify as children's drawings, but not as graphs. Take an engineering book or even a datasheet, something on measuring characteristics of induction motors for example. There should be properly formatted, scaled, labelled and distinguishable graphs made from measurements in there, with measuring points and characteristics clearly distinguishable.

One of my harsher laboratory doctors threatened my college for this kind of buffoonery. Maybe he rubbed off on me, but maybe I just expect to be respected when presented with data. Making colorful lines from point to point does not make it a graph. We simply might have different expectations, that in itself is okay. There are different schools on everything. Rants are not welcome, so I probably should have skipped that commentary. It is just too painful to see the level dropping so fast so low.

Remember, that particular work is older, but mentions aspects of interest. If you look for empirical data, depending on where you live, you could contact a polytechnic or university researchers in the field of semiconductors or something more specific targeting the memory area and simply ask them for information. I bet am certain, they will be glad to help or at least point you in the right direction.

Edit: I do not bet. Also should you get to anything specific, maybe you could share the newly acquired information.

Last edited by emeres (2014-09-19 21:04:28)

R00KIE · 2014-09-19 21:49:11

Regarding the images, in my oppinion they are in line with what you will find in any other website. I agree that there are a few things that could be improved but I've seen worse, in other websites, datasheets and scientific papers. I guess each one of us has something where less that perfect will not cut it

Regarding the work you linked to, sometimes older is better, by this I mean it is simpler to understand or get to grips with something new, as things are probably not as complicated are more recent works that build on older ones.

Currently I don't know of anyone working in the field of semiconductor/memory/flash research (the hardcore low-level stuff) so I would probably search about it on my own. I suppose ieee and others will have papers about this but now I don't have much free time to dig into this. Sharing might be a problem, you probably know as well as I do that publicly sharing ieee papers is frowned upon and I don't want to get into trouble. If the data is freely available then I can share what I find.

Regarding data retention, it seems these guys[1] heard me It's these kind of things we don't see tested very often, in this case it is 1 month, but what about 6 months, 1 year or longer? Not that rotating media isn't starting to get scary too[2] but that's off-topic.

[1] http://www.anandtech.com/show/8550/sams … on-the-way
[2] http://www.ewh.ieee.org/r6/scv/mag/MtgS … tation.pdf

IvanHomeless · 2014-09-20 14:08:55

I've had ext4 with full disk encryption on an SSD for 4 years now with no problems at all. I have another one that's 2 years old and has seen ext4, Btrfs, FFS, and more, plus full disk encryption each time. It's also fine. I think people get freaked out about the fact that SSDs have a limited lifespan, but forget that HDDs will fail as well, possibly sooner than an SSD. It's just easier to be conscious about an SSD wearing out for some reason.

emeres · 2014-09-20 19:19:15

@R00KIE Those are far from acceptable, but like you wrote, completely in line with other websites, which just shows how low the "tech savvy people/industry" got.

R00KIE wrote:

Sharing might be a problem, you probably know as well as I do that publicly sharing ieee papers is frowned upon and I don't want to get into trouble.

Of course its understandable and of course I did not mean that, stay safe. Should I come across anything relevant, I will report back, as long as I remember it that is.

Arch Linux

#1 2014-09-18 11:31:00

Is ext4 bad for lifetime of SSD?

#2 2014-09-18 11:36:19

Re: Is ext4 bad for lifetime of SSD?

#3 2014-09-18 12:01:55

Re: Is ext4 bad for lifetime of SSD?

#4 2014-09-18 12:46:03

Re: Is ext4 bad for lifetime of SSD?

#5 2014-09-18 12:53:36

Re: Is ext4 bad for lifetime of SSD?

#6 2014-09-18 13:16:47

Re: Is ext4 bad for lifetime of SSD?

#7 2014-09-18 14:06:37

Re: Is ext4 bad for lifetime of SSD?

#8 2014-09-18 15:48:14

Re: Is ext4 bad for lifetime of SSD?

#9 2014-09-18 19:09:21

Re: Is ext4 bad for lifetime of SSD?

#10 2014-09-18 20:01:58

Re: Is ext4 bad for lifetime of SSD?

#11 2014-09-18 21:30:19

Re: Is ext4 bad for lifetime of SSD?

#12 2014-09-18 22:36:00

Re: Is ext4 bad for lifetime of SSD?

#13 2014-09-19 02:23:28

Re: Is ext4 bad for lifetime of SSD?

#14 2014-09-19 02:59:11

Re: Is ext4 bad for lifetime of SSD?

#15 2014-09-19 09:09:15

Re: Is ext4 bad for lifetime of SSD?

#16 2014-09-19 09:59:34

Re: Is ext4 bad for lifetime of SSD?

#17 2014-09-19 16:42:09

Re: Is ext4 bad for lifetime of SSD?

#18 2014-09-19 16:56:38

Re: Is ext4 bad for lifetime of SSD?

#19 2014-09-19 19:35:30

Re: Is ext4 bad for lifetime of SSD?

#20 2014-09-19 21:01:52

Re: Is ext4 bad for lifetime of SSD?

#21 2014-09-19 21:49:11

Re: Is ext4 bad for lifetime of SSD?

#22 2014-09-20 14:08:55

Re: Is ext4 bad for lifetime of SSD?

#23 2014-09-20 19:19:15

Re: Is ext4 bad for lifetime of SSD?

Board footer