back to article SSDs in the enterprise: It's about more than just speed

Enterprise solid state drives are gaining traction, but their predominant focus is still performance. The need for speed has driven SSDs into applications where HDDs previously reigned, but for those of us who aren’t high-frequency traders, solid state will need to demonstrate some other benefits. What are they, and how …

  1. Paul Crawford Silver badge

    in the long run, will be about economics and the dollar-per-bit cost

    Generally it always is, as performance vs cost for RAID / short-stroking, etc, has been covered.

    But for now if you have lots of data (e.g. tens of TB) and limited / sequential access patterns HDD is still way cheaper. When that changes we will buy SSD in a flash.

    1. Rainer

      Of course - anything that basically just sits there is a bit of a waste for SSDs.

      But hyperconverged usually means some form of VMs. And they need to be able to move around.

    2. phuzz Silver badge

      We've found that buying a few boxes of SSDs, and the same number of knock-off HP caddies is cheaper than buying HP's own brand SAS disks.

      Of course, you can't get new SSDs under warranty, but by the time you need to they're cheap anyway. Plus, because the SSDs are speeding up the databases so much, we don't need to load balance across as many servers, so there's another big saving there.

      1. TheVogon

        "We've found that buying a few boxes of SSDs, and the same number of knock-off HP caddies is cheaper than buying HP's own brand SAS disks."

        Until you need warranty support for your lost data that is...

        1. Adam JC

          Warranty support for lost data? I don't think I've ever encountered a HDD/SSD manufacturer anywhere that offers to recover data from a failed unit under warranty. That's what backups are for...

          1. TheVogon

            "I've ever encountered a HDD/SSD manufacturer anywhere that offers to recover data from a failed unit under warranty."

            I have had HP in on site before when a RAID controller firmware bug trashed a disk array. If that array had had non HP disks in it I doubt they would have helped.

            "That's what backups are for..."

            Backups don't recover data that changed since the last backup...

        2. phuzz Silver badge

          HP do warranty support for lost data? News to me. We just use RAID (to reduce the likelihood of losing data), and backups (to protect the data when it is, inevitably, lost). Oh, and multiple DB servers for redundancy and load balancing.

          Assume that your disks (or SSDs, whatever) will fail and plan accordingly, don't hope that HP can get the data back.

  2. quxinot Silver badge

    @Paul Crawford

    Thank you. I wasn't sure who sponsored the outdated advertising article I just read, but I'm glad to see that I'm not the only person who thinks in terms of real money being spent.

    SSD's have good and bad about them, as do normal hard drives or even some of the newer tech drives (Helium, etc). Any of them can do the job, but it moves the compromises around depending on what you use and how it's configured.

    1. Anonymous Coward
      Anonymous Coward

      I wasn't that far into the article when I thought, hang on, is this from ten years ago? A little bit further in it was clear that this was by someone who had a lot of SSDs to shift.

      All of those who sell SAN storage devices include both HDD and SSD options, and generally you get both in the same box. Is how they utilise the types of media that distinguish them.

      SSDs are still way more expensive than HDDs. And if you put a physical SSD into a server and only use a fraction of it, generally speaking you've wasted your money.

      Data at rest does not need fast storage.


    The writing has been on the wall for a while now.

    Stationary silicon will replace rotating rust. Despite my early experience with problematic SSDs I'm going all in for the future. No more purchases of HDDs will take place here.

    1. Anonymous Coward
      Anonymous Coward

      Re: The writing has been on the wall for a while now.

      And where is here? Your laptop? Or do you own an enterprise environment with PB of data? And do you have an infinite budget?

      Tape was supposed to have died last century. And lo and behold, people are still buying it.

      1. Anonymous Coward
        Anonymous Coward

        Re: The writing has been on the wall for a while now.

        "Tape was supposed to have died last century. And lo and behold, people are still buying it."

        From a timeline found somewhere on t'internet:

        1972 - IBM begins development on its last tape drive (3480) ever because of the declining cost of disk drives.

        1. Anonymous Coward
          Anonymous Coward

          Re: The writing has been on the wall for a while now.

          I guess that's why people think it must be dead.

          Meanwhile, for people how don't take uncited facts to be true, you can google some other IBM enterprise tape drives from the list below:

          3480, 3480 IDRC, 3490, 3490E, 3590......., TS1120, TS1130, TS1140, TS1150.

          Since the "last drive" which could store 200 MB per cartridge you can now do 10 TB per cartridge (50K improvement since the last tape drive invented) and that's before what the lab guys and girls are working on (in theory 220TB cartridge as a tech demonstration in 2015).

          Obviously LTO never happened either. 7 times so far.....

  4. Anonymous Coward
    Anonymous Coward

    SSD's also saved jobs....

    ... For the first time - spending a moderate amount of $$$ - delivered a noticable performance increase all the way up to the application while hiding all the architectural sins of your predecessors (at least for the time being).

    IT managers could finally shine in front if their business stakeholders while regurgatating terms such as "disruptive technology", "agility" , "time to market" - or whatever their storage vendor dreamt up.

  5. Stuart 22 Silver badge

    Don't forget the CPU

    The article does point out that the power drain of a system is rather more than than storage alone. However the speed of SSD can ironically allow the deployment of less powerful CPUs and still achieve the same overall performance. With a less powerful CPU one may need a smaller PSU with fewer losses which in return require less cooling to create a virtuous circle in overall power requirements.

  6. Efros

    And of course

    You can play Open Cities mod in Skyrim with affecting your FPS too much.

  7. Dazed and Confused

    Not all about performance

    I'm sure I remember reading an article here a few years ago where someone from Facebook was talking an a semiconductor conference about there needs for storage. He put forward a use case for SSDs saying they had vast qualities of data that was effectively write once, read hardly ever. All those millions of photos of peoples cats they posted and no one looks at after the first few minutes. They were after being able to save all the power that went in to keeping all that rust spinning.

    The upfront cost of SSDs is very visible, but for many customers it is more difficult to quantify the cost saving from the power consumption. BTW the figures you quote don't seem to make sense. You talk about a 90% power reduction but only a 30% heat output reduction, surely these figures need to match. Cutting down on the heat out also reduces the power bill from the AC.

    1. Anonymous Coward
      Anonymous Coward

      Re: Not all about performance

      HDDs don't spin when they arn't being accessed. For archival storage HDDs can use far less power than SSDs over long timespans. Matching the technology to the usage beats arbitrarily picking one or the other and using it inappropiately.

      1. Peter Gathercole Silver badge

        Re: Not all about performance @AC and spinning HDDs

        Whilst it it true that there are storage systems that do spin down disks when they are not in use, this is not the norm. The problem with spinning disks is that frequently, the most likely time they fail is when they are spun up after being spun down for a period of time.

        I recently was in charge of an estate with ~5000 disks in it. It ran 24x7, and was only powered down if there was work being done on the site UPS. When we powered it back up, we inevitably had problems with some of the HDDs, leading to disk-recovery procedures needing to be invoked.

        I know that in the archival cold-disk vaults, they end up with a complex RAID/mirroring environment such that data was stored on *several* hard disks, so that when the inevitable failures happen, the data is still safe. You also have to cope with bit-rot.

      2. Gerhard Mack

        Re: Not all about performance

        "HDDs don't spin when they arn't being accessed."

        Do you really think the user will tolerate waiting several seconds for the drive to spin up when you are accessing old pictures on Facebook? The page will just hang for several seconds before it starts to load.

        The advantage of SSD in this case, is that it has almost no startup time.

        1. Anonymous Coward
          Anonymous Coward

          @Gerhard Mack - several second delay when accessing old pictures on Facebook

          Actually, that's EXACTLY how Facebook handles cold data, and is why it generally does take several seconds when pulling up an old picture on Facebook that hasn't been accessed by anyone in some time...

          1. Dazed and Confused

            Re: @Gerhard Mack - several second delay when accessing old pictures on Facebook

            Here's the link to the El'Reg article


    2. Anonymous Coward
      Anonymous Coward

      Re: Not all about performance

      The heat reduction shouldn't match power reduction at all, otherwise you would be getting the spin in hard drives for free!

      1. Dazed and Confused

        Re: Not all about performance

        > The heat reduction shouldn't match power reduction at all, otherwise you would be getting the spin in hard drives for free!

        Any energy you put in to spin up the disk at power-up is then dissipated as heat at power-down. Any energy you pump into the disk to keep it spinning is dissipated as heat due to friction losses. A tiny amount of the energy will get propagated out as sound waves. The rest just goes as heat.

        > HDDs don't spin when they arn't being accessed.

        Most do. All of mine do.

        The spin up time is way too long to want to wait for. In archival situations where a disk array is being a replacement for a tape library then you might be OK with waiting for the damn things to come up to speed and stabilise but in most online data access cases they need to spin the whole time. Most enterprise class disks even regularly wobble the heads around to keep them fit as well.

        Most failures happen when disks are spun up or down.

  8. David Austin

    SSD Burn out

    How does the known end of life date of SSD's caused by Flash Degrade (As opposed to a "Perfect" hard drive running indefinitely) factor into the TCO at the Enterprise level?

    1. doowles

      Re: SSD Burn out

      SSD burnout is a non issue

    2. Gerhard Mack

      Re: SSD Burn out

      Given the storage array I'm in the process of retiring (10 year old drives) and it's drive failures every other month, and the regular age related deaths of storage on our existing servers, "perfect" does not exist in the hard drive world.

      My newer servers and SANs have a write countdown that tells me when the SSD is wearing out and that should give me a far more predictable. "drive is wearing out, replace soon before it becomes a problem" rather than "dead drive, the RAID is now degraded. lets rush out and buy a new one" Or the worse yet "two drives are dead with no warning and we have lost data"

      Unpredictable failures have a cost that should be accounted for in any TCO calcualation.

      1. Anonymous Coward
        Anonymous Coward

        Re: SSD Burn out

        The nice thing about NAND failure is that it fails in a way that stops it from being erased & rewritten. You don't lose data when a NAND device wears out, you just can't write to it anymore. If only hard drives failed in such a friendly manner.

        Now the controller may not always handle these scenarios correctly - if you look at storage report's long term testing some controllers will lock up at some point (probably when there are no free blocks it is able to write) because that's a difficult condition to test for - their test took ages to complete. But over time most of those bugs will be found and fixed, and you'll see graceful degradation.

        Of course that's not a problem in the enterprise world, because you have something monitoring SMART and letting you know well in advance of your SSD's failure. SMART often warned of hard drive failure, but sometimes it would just fail without any warnings first because hard drives did not fail in such a predictable and graceful manner.

        Obviously the electronics can go bad in a flash drive, if the controller takes a dump you're screwed and won't have any advance warning. But hard drives are no different in that respect.

  9. doowles

    I've got an 8tb SSD based NAS back home, been using it for over a year now.

    The days of large storage on SSD are here.

    1. Anonymous Coward
      Anonymous Coward

      You could have bought one using HDDs for a fraction of the price.

  10. jms222

    SSDs do NOT enable de-duplication as the article states.

  11. Anonymous Coward
    Anonymous Coward

    spinning rust

    Spinning rust is about 18 months from extinction. Good riddance.

  12. DeepStorage

    Power is small beans

    a 10TB hard drive (cost $200) uses <200Kwh of power per year. At $0.20/kwh (high end of US cost) and PUE of 2 (1Kwh of AC for each Kwh of gear) that's $80/year of power. The equivalent (capacity, ent feature set) SSD Samsung's PM863 3.84TB costs $2000. Even if it uses no power at all the payback is MANY years.

  13. waltercarroll

    SanDisk’s Extreme Pro is a direct successor to the Extreme II (see below). Just like its predecessor the drive lives up to the expectations implied in the model name. It is slightly ahead of the Extreme II in most areas with sequential read speeds of 550 MB/s and write speeds of 520 MB/s (4K random read/write 100K/90K IOPS), although you will hardly notice the difference in everyday tasks.

    More importantly though, SanDisk is confident enough to offer a 10-year warranty with the Extreme Pro – a unique offer in the consumer segment. The 19nm MLC NAND is allegedly good for writing 22 GB of data per day for 10 years. Consequently, when this drive finally wears out in the average system built today, the SATA interface will be long since obsolete.

    1. FrankAlphaXII

      Its not entirely unique, rare yes, but they don't have a monopoly on a 10 year warranty for a consumer SSD. Samsung have the same thing on their 850 Pro, which is the reason I coughed up the extra 40 bucks for the pro instead of the evo.

      I'd be amazed if SATA is still used at all in 9 years, but its nice to have the warranty regardless.

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon

Biting the hand that feeds IT © 1998–2021