Using SSDs with "disk thrashing" applications?

Discussion in 'SSDs & Data Storage' started by athenian200, Jul 14, 2014.

  1. athenian200

    athenian200 Gawd

    Messages:
    727
    Joined:
    Mar 29, 2012
    I've heard that SSDs don't last as long as HDDs when you're dealing with write-intensive operations. I really like the speed of SSDs on my gaming computer, but I worry about their longevity in more demanding write scenarios.

    I have a Linux machine that I like to play around on. I'm constantly compiling applications and reinstalling operating systems on it. With computers I use this way, my HDDs last 2-3 years (compared with 5-6 years on "normal" machines). I know it's pretty tough on disks in general, because the drive light is constantly on.

    I just wore out another hard drive, and I'm considering upgrading to an SSD. The thing is, this would be the only drive in the computer. A lot of people say you need a mechanical and an SSD, but I don't really want to mess with a dual configuration.

    Would an SSD be able to last the same 2-3 years in the scenario I described, with constant compiling/reinstalling stuff? Or would it wear out a lot more quickly than an HDD?

    I basically want to pay a little more for the extra speed, I don't really expect it to last longer than 3 years given what I'm doing. I just want to make sure it won't die within a year under my brand of hard disk abuse.

    I haven't heard a lot of reports of how well SSDs hold up in my specific situation.

    Any advice?
     
  2. munkle

    munkle [H]ardForum Junkie

    Messages:
    11,196
    Joined:
    Jan 16, 2005
  3. daglesj

    daglesj [H]ardness Supreme

    Messages:
    5,117
    Joined:
    May 7, 2005
    Buy one, thrash it, then buy another for half the price and twice the size of the existing one if/when it fails.

    Simple.
     
  4. evilsofa

    evilsofa [H]ardForum Junkie

    Messages:
    10,078
    Joined:
    Jan 1, 2007
    This is the basic question you need to answer: how much data do you write to the disk each day? We're talking about the average long-term daily writes;
     
  5. Meeho

    Meeho [H]ardness Supreme

    Messages:
    4,472
    Joined:
    Aug 16, 2010
    Really, this. I'm not sure where all this SSD pampering is coming from. It's as if people are told to buy them, but don't use them or they will explode. Like removing page file from SSD, which is the stupidest advice ever since it's the kind of use they are made for.

    As for OP's disks lasting only 2-3 years, I don't think that is normal. Something else is killing them.
     
  6. Red Squirrel

    Red Squirrel [H]ardForum Junkie

    Messages:
    9,211
    Joined:
    Nov 29, 2009
    The kind of thrashing you describe is probably not enough to be an issue. It's only really while you are using it right, or are you leaving processes that do stuff over night or 24/7?

    It's not normal that your HDDs are not lasting though, HDDs don't really wear out, they just randomly fail. Are you doing anything that causes them to not shut down properly? That is bad for HDDs.
     
  7. athenian200

    athenian200 Gawd

    Messages:
    727
    Joined:
    Mar 29, 2012
    That sounds like a lot of data. I doubt I would use them up as quickly as they did in those tests.

    That's basically what I've been doing with HDDs.

    Hmm... well, I've never sat down and measured it. I think it's probably 10-20GB per day. Not sure how much impact reformatting the drive would have on that number, though.

    Yeah, the idea people have been giving me is that HDDs are slower but more reliable, while SSDs are built for speed but will die quickly.

    That's NOT normal? All I know is that my optical drives fail even more quickly. If I get a new computer or a video game system, I can count on the optical drive failing within 1-2 years of heavy use. The way things generally go for me is that the optical drive fails first, then the hard drive, then the power supply. Other parts of my computer pretty much never fail.

    I know it's the drive that's failing on the consoles... my SNES and N64 which lack these drives still work. My PS2, PS3, Gamecube, and Wii, all had to be replaced a few times in their lifetime just because of the drive failing, and the manufacturers even confirmed it. They always had to send me a new system under their 1 year warranty, at least the first time.

    Not intentionally, but I do have power outages in the summer, and sometimes solid lockups that force me to cut power.

    I leave them on overnight, but mostly idle. I let the computer update itself, have it give me alerts if someone contacts me on Skype, etc. But it doesn't do anything particularly heavy while I'm asleep.
     
    Last edited: Jul 14, 2014
  8. NetJunkie

    NetJunkie [H]ardForum Junkie

    Messages:
    9,682
    Joined:
    Mar 16, 2001
    You won't kill a good modern SSD doing any of that.
     
  9. drescherjm

    drescherjm [H]ardForum Junkie

    Messages:
    14,525
    Joined:
    Nov 19, 2008
    This is very wrong. In most cases you have to go out of your way to try to kill an SSD. Although there have been reported cases of rogue applications that write continuously 24/7/365 that have been reported to wear out an SSD.

    Get an SSD from a reliable manufacturer like Samsung, Crucial or Intel. And do not get the smallest drive possible. These days I would not bother with drives smaller than 240GB. For wear leveling purposes a larger SSD will allow you to writer more data per day.

    If you are utlra worried get a Samsung 850 pro which has a 10 year warranty.
     
    Last edited: Jul 14, 2014
  10. Red Falcon

    Red Falcon [H]ardForum Junkie

    Messages:
    10,017
    Joined:
    May 7, 2007
    Actually, SSDs can die from too many cold/hard poweroff sequences (power loss).
    I've seen this personally, and a few articles have been written online about it as well.

    The one consumer SSD known to withstand this happening is the Intel S3500:
    http://beta.slashdot.org/story/196131


    An UPS would be simple solution to this issue, however.
     
  11. drescherjm

    drescherjm [H]ardForum Junkie

    Messages:
    14,525
    Joined:
    Nov 19, 2008
    I assume death in this case is because the mapping table gets at least partially corrupted and the firmware detects this and just locks the user out of their data.
     
  12. Meeho

    Meeho [H]ardness Supreme

    Messages:
    4,472
    Joined:
    Aug 16, 2010
    20 GB a day for years shouldn't be a problem.


    Looks like you have some power issues. That sounds like far too many failures. I would look into getting an UPS and a good quality PSU if you don't have one.
     
  13. kumquat

    kumquat [H]ardness Supreme

    Messages:
    5,269
    Joined:
    Dec 7, 2005
    Doesn't matter how long it'll last. SSDs are so much faster in the "disk thrashing" arena, and fail in such a more predictable manner, that it totally obviates lifespan.
     
  14. athenian200

    athenian200 Gawd

    Messages:
    727
    Joined:
    Mar 29, 2012
    400-500GB is about the capacity I buy drives at now, and it has been since 2009. I don't even fill those drives up, but they stopped making smaller hard drives a while back.

    Well, as long as they don't die from power loss any faster than HDDs, then it doesn't matter. Corruption isn't an issue, I can afford to lose the data. As long as the drive can be reformatted and used, it's good enough for my purposes.
    You would think so, but when Verizon came in and installed my Internet, the tech told me I wasn't allowed to use my UPS with anything directly attached to their hardware. They said that if I wanted to use Ethernet with their router, I had to plug my computers into the wall. I have to exclusively connect with Wi-Fi or upgrade to business-grade FiOS in order to be allowed to use a UPS. Yet another reason I miss AT&T, but that's another topic by itself.

    I was using the UPS on my laptop for a while so I would have extra redundancy on that, but it eventually died.

    It seems like this whole power protection mess is very controversial and expensive. :(
     
    Last edited: Jul 14, 2014
  15. drescherjm

    drescherjm [H]ardForum Junkie

    Messages:
    14,525
    Joined:
    Nov 19, 2008
    For SSDs you can get 64 GB and 128GB current generation drives. I would avoid these for performance and endurance reasons.

    For hard drives. With hard drive makers using 1TB platters now for the 3.5in form factor, I can understand 500GB drives with 1 head but I am not sure smaller drives are just old stock or what.

    I know a person in FIOS tech support I'll see if I can get a clarification of this seemingly ridiculous regulation.
     
  16. mwarps

    mwarps [H]ardness Supreme

    Messages:
    7,006
    Joined:
    Oct 6, 2002
    That's a steaming pile. There's no legitimate technical reason for this. Buy a UPS and be done with it.

    Plugging your hardware into the wall is like screwing a prostitute without a condom in Thailand.
     
  17. drescherjm

    drescherjm [H]ardForum Junkie

    Messages:
    14,525
    Joined:
    Nov 19, 2008
    Maybe they do not want you to plug their modem into your UPS possibly because cheap UPSs do not have sine wave output. I can't any valid reason why they would want or require your PC to be connected to the wall though.

    I would be very concerned if there was no surge protection.
     
  18. Red Squirrel

    Red Squirrel [H]ardForum Junkie

    Messages:
    9,211
    Joined:
    Nov 29, 2009
    Yikes yeah I would definitely get a UPS even a cheap one. Hard shut downs are hard on the whole computer, but especially the storage device. That's probably why your drives are dying so fast.

    That loud click you hear when the power goes out? That's the head slamming against the platters. Though I've heard that some drives actually use the centrifugal force of the spinning platter to generate enough electricity to properly park the head, but I don't know if all drives do this, probably not.
     
  19. athenian200

    athenian200 Gawd

    Messages:
    727
    Joined:
    Mar 29, 2012
    Well, they said an ordinary surge protector or power strip was fine when I asked. When they made me disconnect my UPS, I pulled out my surge protector and asked if I could use that instead. They said anything without battery backup would be okay.

    So everything is connected to a surge protector power strip thing that doesn't have battery backup. I really didn't want to plug my stuff directly into the wall, so that was the compromise we came up with.

    I just tend to refer to surge protectors as being a "direct/wall plug" connection because I use them so often that I think of it as being a default connection, plus everyone says it's basically the same as having it directly connected.
     
    Last edited: Jul 14, 2014
  20. cbf123

    cbf123 n00b

    Messages:
    51
    Joined:
    Dec 12, 2013
    The only reason I can think of why they might be concerned about this is if their modem/router is susceptible to interference from a cheap battery-backed UPS. A low-end UPS is going to have a stepped approximation to a sine wave, which means lots of extra high-frequency noise compared to a pure sine wave. It's possible (but not likely) that this noise get onto the ethernet cable and interfere with the router. More likely they just don't want to deal with any tech support issues.

    Personally I'd have no problem just putting the computer(s) back on the UPS.
     
  21. athenian200

    athenian200 Gawd

    Messages:
    727
    Joined:
    Mar 29, 2012
    It probably was that, they mentioned something about electrical noise.

    The UPS systems that put out a pure sine wave were way out of my budget when I last checked, though.

    I'm looking around at them now actually, and I'm a bit disappointed at the range available. There are these really cheap ones that only output 450W, or the high-end ones that output over 1000W. A 450W supply would have been enough to run my old computer, but I had to put a 650W power supply in the new one in order to run my GTX 580 comfortably.

    Does the watt rating of the UPS need to be at least as high as the watt rating of my PSU? Or can it be lower? The UPS devices that appear rated to run my system cost 3x more than replacing the hard drive every 2 years...
     
    Last edited: Jul 14, 2014
  22. FLECOM

    FLECOM Modder(ator) & [H]ardest Folder Evar Staff Member

    Messages:
    15,592
    Joined:
    Jun 27, 2001
    I use an old OCZ Vertex 2 120gb SSD for the scratch directory of my usenet downloading box, thing moves gigs in and out unparing/checking etc... drive has been doing this for a couple years now without issue
     
  23. Xaero_toast

    Xaero_toast Limp Gawd

    Messages:
    391
    Joined:
    Dec 29, 2008
    I thought I was done with worrying about a hard drive failure when I got my first SSD, and within a year, it had what I can only describe as a "damned odd failure".

    It was a corsair F120. It would become nonresponsive at a predictable, repeating time interval past power on. Memory fails me exactly what it was, but it was in the neighborhood of 90 seconds. It was just long enough to get through POST, load windows and login, and have about 30 seconds of time at the desktop. When the time was up, the drive became unresponsive.

    I hot swapped it into another system, so I could use the full 90 second window to copy off files, without waiting for bootup. I recovered all my files, RMA'd the drive, and a week later had a refurb unit back. It's been flawless since then.


    If someone really wanted to murderize an SSD, they'd put it in a Tivo.
     
  24. Red Squirrel

    Red Squirrel [H]ardForum Junkie

    Messages:
    9,211
    Joined:
    Nov 29, 2009
    Honestly I would just put everything back, it's the first time I hear of an ISP saying not to put the modem on a UPS. So each time your power goes out you lose internet? Seems kinda odd they'd want that. The average user is going to call tech support not realizing that the modem is off because the power is out.

    Pure sine UPSes are huge bucks and typically only found in large scale commercial applications. You don't have to size it as high as the PSU provided you're not drawing that much. Keep in mind that UPSes are sold and rated in VA's while everything else is watts.

    1000VA = 600w or so.
     
  25. FnordMan

    FnordMan [H]ard|Gawd

    Messages:
    1,727
    Joined:
    Apr 22, 2011
    Have a look at either a refurb APC Smart UPS (i've heard good things about refurbups.com) or a middle range Cyberpower one. Both should have pure sine wave output and may even make power better than the power company, depending.

    Far as rating goes: a 650W PSU doesn't always use 650W. Heck, it can use more if it's near full load. If you're not sure pick up a cheap Kill-a-watt and plug the machines in question into it then crank up the load (CPU and GPU) to get an idea of what the power usage is. (then add 25-50% or so and shoot for that)
     
  26. zrav

    zrav Limp Gawd

    Messages:
    163
    Joined:
    Sep 22, 2011
    SMART can only predict about 30-40% of HDD failures and they can happen rather suddenly.
    SSD failures other than wear-out are very rare (not counting specific models that had issues). And wear-out is monitorable via SMART and very predictable.
     
  27. athenian200

    athenian200 Gawd

    Messages:
    727
    Joined:
    Mar 29, 2012
    Oh, if I were USING 650W, I would have bigger PSUs. I'm actually drawing about 400-450W under full load. However, it's a lot less than that when the graphics card isn't stressed. Still, there's no guarantee I won't be playing a 3D game or otherwise GPU computing when a power outage happens. Probably better to assume I will be.

    I mainly wanted a 650W UPS in order to have a bit of cushion. A 450W device would be just barely enough for the computer, assuming I don't plug my monitor or printer into it.

    I am finding some better ones at better prices now, looks like I may just be getting a good UPS after all.

    I wonder if people just think SSDs are less reliable because their writes are limited? It's strange, pretty much everyone agrees that SD cards and Flash drives are more reliable than floppy disks, but no one seems to think SSDs are reliable compared with hard drives. Even though the same basic technologies are involved.

    I've been looking at Crucial's site, and it seems they have a 3-year warranty on their drives. So clearly they expect it to last at least that long. Companies rarely put that kind of warranty on something that has a high failure rate.

    With an HDD, I generally know it's started going bad when I start hearing scraping metal and a longer "dragging" sound when reading. If I let it go from there, I start getting seek errors and bad sectors regularly. I've been through that enough times to know when I need a new one, I rarely wait for a total failure. What's SMART? Some kind of disk health monitoring utility?
     
    Last edited: Jul 16, 2014
  28. FnordMan

    FnordMan [H]ard|Gawd

    Messages:
    1,727
    Joined:
    Apr 22, 2011
    Self-Monitoring, Analysis and Reporting Technology. As stated though, it's rarely effective at warning for a drive failure. I've had a couple that were just *bam* dead. Controller failure in one and I forget what happened with the other.
     
  29. Xaero_toast

    Xaero_toast Limp Gawd

    Messages:
    391
    Joined:
    Dec 29, 2008
    The bam will do it every time. ;)
     
  30. HammerSandwich

    HammerSandwich [H]ard|Gawd

    Messages:
    1,116
    Joined:
    Nov 18, 2004
    Sounds like an overheating component.

    You could get by with a $120ish Cyberpower PFCLCD, which is pure sine. Consider saving a little more for a bigger model. Bigger is better with a UPS, runtime improves & you'll be able to plug other stuff into it, especially with sine-wave.
     
  31. Xaero_toast

    Xaero_toast Limp Gawd

    Messages:
    391
    Joined:
    Dec 29, 2008
    It could have been, but I wasn't going to open it up while it was warranty.

    The thing is, it didn't really need a cooldown. It could be power cycled, and it would come back immediately, and run for the exact same duration of time.

    Thank goodness SSDs read so quick. Can you imagine recovering a standard hdd 90 seconds at a time?
     
  32. Argentum

    Argentum [H]Lite

    Messages:
    89
    Joined:
    Nov 14, 2006
    There's been a lot of chatter for years but few do the real world analysis- which would include understanding your own usage needs.

    With SD cards and flash drives and I think it's just expectations - how much data in absolute terms really gets written to your typical USB drive in a given time period? These will wear out too, probably at faster rates than SSDs, but practically no one ever hits the limit before the key is broken, lost, or otherwise obsoleted.

    Personally, I've had so many USB keys go corrupt on me for one reason or another that probably has nothing to do with wear that I never use them for anything other than transient file transfers where the original data is elsewhere. My digital cameras today must have WiFi that transfers my photos to the cloud or home network ASAP.

    Anyway, here's what I hope is my value add for this thread:

    I have a home lab with an SSD array of 6 drives in a Windows Storage Space pool with multiple VMs on it, including many that run databases 24/7 (for System Center products - SCCM, SCOM, SCVMM, SCDPM).

    The worst suffering SSD in that pool is a 256GB Samsung 840 Pro showing a 56% SMART health rating after 349 days involving 10.71TB of writes. If I take the SMART health rating verbatim and assume the drive is dead when it hits 0% (which is likely not going to be the case based on the previously mentioned TechReport test implying that the SMART rating is overly conservative), the drive is still good for another 1.25 years. Realistically it'll probably go much longer. The other drives (also 840 Pros) have held up better for one reason or another.

    That drive cost me about $240 originally. I expect to be able to replace it with the equivalent or better (i.e. TLC NAND and/or 2x capacity) for at least 20% less than that price in 1.5 years. If I really cared enough about making each drive last longer I could expand the pool and lower the average write rate sooner.

    Maybe my expectations are wrong. Buying 2TB drives for $60 in 2011 I also expected to be able to get 8TB drives in 2014 for under $100 for my mass archival storage needs, but it didn't work out that way. But IMO, there is no technology that isn't "consumable" over time so I see everything - even RAM and CPU in an era where advancements have slowed compared to prior decades - in terms of TCO per abstract applicable units over time (e.g. $/GB/year) rather than upfront costs.

    Ultimately, one should expect failures at a certain rate and understand that the random nature of the universe makes those failures sometime happen sooner and sometimes later than the statistical average. For me, aside from redundancy (which is more for data availability rather than data integrity/retention), planning for those failures means backing up my data frequently :)