Linux - NewbieThis Linux forum is for members that are new to Linux.
Just starting out and have a question?
If it is not in the man pages or the how-to's this is the place!
Notices
Welcome to LinuxQuestions.org, a friendly and active Linux Community.
You are currently viewing LQ as a guest. By joining our community you will have the ability to post topics, receive our newsletter, use the advanced search, subscribe to threads and access many other special features. Registration is quick, simple and absolutely free. Join our community today!
Note that registered members see fewer ads, and ContentLink is completely disabled once you log in.
If you have any problems with the registration process or your account login, please contact us. If you need to reset your password, click here.
Having a problem logging in? Please visit this page to clear all LQ-related cookies.
Get a virtual cloud desktop with the Linux distro that you want in less than five minutes with Shells! With over 10 pre-installed distros to choose from, the worry-free installation life is here! Whether you are a digital nomad or just looking for flexibility, Shells can put your Linux machine on the device that you want to use.
Exclusive for LQ members, get up to 45% off per month. Click here for more info.
The "dirty little secret" is that all SSD's will slow down and it won't take too long to notice it on tests.
No they won't...not with TRIM, that's the point.
The 3-year old 256 GB SSD in my old laptop is still just as fast as it was on day 1. Spec'd at 500 MB/s read and 260 MB/s write, I get 422/286 in my tests, which is about what I got when it was new.
+1 to suicidal eggroll. Theres plenty of life in any modern SSD with or without journaling. Ssd's are NOT overgrown jump drives. They do load balancing, wear leveling,and lots of other esoteric stuff because the producers know evrybody's hand wringing over the finite write life and they want the technology to succeed.
I will disagree in one aspect. Even without journaling, you probably won't trash your data in an unclean shutdown. The data's less secure, but the increased risk is similar to the lifespan increase of journaled v unjournaled.
In thr dark ages we crashed much more than we do now. I lost data like everyone else, but it was by no means a foregone conclusion. You killed a chicken, prayed to the moonand crossed fingers.
Usually it worked. If not, you spun the tape and cursed a lot
I have 20 year old mechanical drive that have been in an industrial environment. They have proven to be worth the money.
I still say I like SSD's but they have a long way to go in order to compete with the best in terms of lifespan. Talk to me in 20 years. My SCSI drives may still be in service. My SSD's won't be I can bet on that. I doubt yours will be working either.
I use SSD's and I use them just like any mechanical drive. When they either wear out or fail or just become too small, I'll get what is new.
Just because people are selling SSD's doesn't mean they have the millions of hours of real world testing behind them. Mechanical drives have had decades and decades of real work on them.
Saying a SSD is the best ever is like saying Linux can't get any malware.
Distribution: Debian Sid AMD64, Raspbian Wheezy, various VMs
Posts: 7,680
Rep:
To counter that I've had, and known of, hard drives fail within the first year and I come across many at work which have failed within the first 5 years. I think in a lot of cases it might be the drive controller rather than the spinning rust itself but it's still fatal.
I do agree though that SSDs ought to be used like spinny disks -- otherwise, to my mind, there's little point in using them.
Personally, in my home PCs, I wouldn't really trust any drive over 5 years old regardless of type.
Yes, spinning rust is likely to last longer and I would still archive to it over SSD but for day-to-day use I don't know.
The golden rule is testing. Theory and "general knowledge" only takes you so far when you really need facts. Moreover, someone ELSE'S testing is worthless to you, unless they test YOUR use case and platform.
I tested (on hardware not my own) several file systems under RHEL7, encrypted and unencrypted, and for my server applications EXT4 was fastest on both rotational and non-rotational drives. The fastest drives were the newest SSD SCSI drives on smart HP controllers with no more than six drives per controller (at 6 the bottleneck of SSD I/O and controller channel I/O reached the corner point of diminishing returns). PCIe and SATA did not approach the same speeds until you had the special case of a single drive per controller, but the SATA and PCIe hardware was not exactly state-of-the-art. Those results MAY be valid ONLY for me, but FWIW...
Due to the performance degradation over time for SSD and rotational drives, I cannot predict that this would hold true over the life of the platform. I did not test F2FS. XFS and BTRFS could not keep up, but BTRFS had some interesting factors: I expect great things from those people down the road! Everything else was slower or worthless.
Let me explain worthless, because it is not that they were BAD. EXT2 was actually fastest, until it crapped out due to the number of files and folders my test generated: hit the limit and bounced! JFS2 was interesting, often on par with XFS but with different performance characteristics, but could not keep up with EXT4.
All things considered (without F2FS considered), I would go with RAID10 SSD using EXT4 with NOATIME and DISCARD for reduced writes and trim support. This is solid, well tested, well supported, and very fast. If we want to go cutting edge, research the best way to implement F2FS storage, but expect some bleeding.
Let me add that memory makes a difference. All of the disk I/O must be buffered in memory, and the more memory (and the FASTER memory) the closer your I/O will approach theoretical maximum. My database application about capped with somewhere around 320G of ram, but that is a SERIOUSLY atypical case. Just be aware, filesystem format, disk hardware , controllers, and CPU are all important, and other factors also influence performance both directly and indirectly. If your machine is a DOG, SSD will only help make it a slightly faster DOG.
Down the road a bit, I expect SSD specific controllers optimised for F2FS or another non-rotational optimizing format. The same might happen with such controllers and BTRFS specific optimizations. At that point the questions will change, and we get to start a whole new round of testing. Can hardly wait!
I still say I like SSD's but they have a long way to go in order to compete with the best in terms of lifespan. Talk to me in 20 years. My SCSI drives may still be in service. My SSD's won't be I can bet on that. I doubt yours will be working either.
I disagree entirely. I have a lot of experience with both mechanical drives and SSDs (as much as possible considering when they came out). All of my consumer-grade mechanical drives, regardless of brand or size, have about a 10-15% failure rate within the first 5 years. This is data accumulated over the last decade from hundreds of drives. Whenever I build a system with mechanical drives, I can practically guarantee it will lose at least 1-2 drives out of every 10 within the first 5 years. Meanwhile out of all of the SSDs I have experience with, dating back over 6 years with millions of accumulated run-time hours, I have only lost one. One, out of approximately 40 drives, the longest of which were turned on in 2009 and haven't been turned off since. I've lost at least 4x that many mechanical drives by percentage in the same time frame.
I'll agree that enterprise level ssd from Intel may be OK. The PCI-e SSD's that were touted to be fantastic almost all failed. $4000 down the drain.
There will always be products which fail, regardless of technology. What sprung to my mind reading the linked aritcle was this: https://en.wikipedia.org/wiki/HGST_Deskstar
Spinning hard drives were a known, proven technology then but it didn't stop a manufacturer making ones which didn't work.
I'm not surprised the Samsung 840s are having issues. They're one of the only SSDs that uses TLC cells, which makes them cheaper, slower, and significantly less reliable than their MLC alternatives (read: basically everything else on the market).
Samsung claimed to have overcome, or at least significantly reduced the endurance penalty of TLC, but I'm not surprised at all that their first attempt at it wasn't a complete success.
They were claimed by almost every tech site as the best of the breed. Best bang for the buck and such.
The company that made the original Revo style drives went under.
The best I idea I ever saw was the card that used common ram and just keep it alive with a rechargeable battery.
We are way off topic however. The OP asks about a filesystem. I still think they ought to use what they want. There are a number of filesystem choices but when it comes down to it, I figure the SSD will either survive intact after some amount of time or fail totally. When it fails a new SSD or card based drive will be on sale.
LinuxQuestions.org is looking for people interested in writing
Editorials, Articles, Reviews, and more. If you'd like to contribute
content, let us know.