Fixing Modern SSDs for the Enterprise

Are modern solid-state drives ready for the enterprise situations?  We've discussed how the market is beginning to combat the high price-to-performance ratios for solid-state technology.  But that doesn't mean that SSDs are ready, technologically speaking, to leap into business servers the second the price-point hits acceptable levels.  At least, that's the stance that HP is taking, predicting that solid-state technology won't see the light of day in the enterprise world until at least 2012.

In an interview with InternetNews.com, Jimmy Daley, an ISS Marketing Manager at HP, said that solid-state drives currently lack a number of key criteria that make them useful in an enterprise environment.  It's unrealistic to think that consumer or even current enterprise-grade SSDs could just be thrown into a server as-is.  That's not to say that companies haven't tried. HP itself uses solid-state drives in its own ProLiant servers.  But this is a specialized scenario, says Daley--modern solid-state drives just aren't ready to hop into data centers for a few of the following reasons:

Lacking Critical Enterprise Features

Solid-state drives don't contain the features of modern magnetic storage that makes them convenient for an IT environment, like the ability to be used in hot-swap installations.  The code for hot-swapping was never build into typical solid-state firmware, and this is a crucial component of enterprise IT environments.  That said, SSDs still take the prize for their energy-savings capabilities over conventional magnetic storage.  

Lack of Performance

It's common-knowledge that SSDs can deliver powerful read performance compared to conventional hard drives. After all, reading data from flash cells is a near-instantaneous process compared to the milliseconds it takes a drive head to move over to the appropriate section of a spinning platter.  But due to their need to erase blocks of flash memory before all writes, SSDs are still underwhelming when it comes to adding new data to the drive.  Until this improves, solid-state technology will still take second places to magnetic-based storage.

Close to Death

Depending on their construction, SSDs can take a hit in their longevity.  While single-level-cell SSDs have been estimated to hold a lifespan of approximately 100,000 uses, multi-level-cell SSDs compact more bits onto each individual flash cell to improve capacity and lower costs.  The downside?  Their total available writes can drop to the thousands range.

But that's not all.  According to a different report by Gerson Lehrman Group, two more factors are preventing widespread SSD adoption in the enterprise market.  The first, cost, is a familiar tune to most solid-state-related articles.  SSDs simply don't offer the capacity that a business could otherwise purchase with magnetic storage.  While the aforementioned nod to power-savings does help the equation, perhaps even allowing a business to consider a hybrid magnetic/SSD environment for different parts of a data center, the waiting game continues for equalizing the price/performance question of modern solid-state technology.

And were SSDs to even become financially viable, stacking a server full of these drives would completely overwhelm modern disk controllers, says the article.  Current controllers, optimized for magnetic storage, wouldn't be able to handle the throughput of a full solid-state array.

It's a big list of issues that confronts solid-state adoption in the enterprise market.  In fact, many of these arguments have been made for consumer-grade drives as well.  There's no question that solid-state technology offers tangible rewards in certain use cases, but expect to be waiting a few more years before the technology pushes their appeal into the mainstream.

  • bustapr
    As long as it saves more energy and doesn't annoyingly crap up when you touch it from the bottom like hard drives , i'll take the ssd.
    Reply
  • TheFace
    Depending on their construction, SSDs can take a hit in their longevity. While single-level-cell SSDs have been estimated to hold a lifespan of approximately 100,000 uses, multi-level-cell SSDs compact more bits onto each individual flash cell to improve capacity and lower costs. The downside? Their total available writes can drop to the thousands range

    Could you be a little more vague and wrong with this statement? 100,000 uses? Is that reads? No. Write cycles? Yes.

    Also, your supposed BIG NEWS is that the Enterprise market is slow to adopt something? OMG what a shocker! Prudent business practices would dictate that you don't jump on the bandwagon of the new hotness (be it whatever, technology etc.) Until it's somewhat proven. SSDs still have a way to go in this respect. Additionally they're becoming much faster with each generation, as well as larger. The market will adapt, but it will take time.

    There's no question that solid-state technology offers tangible rewards in certain use cases, but expect to be waiting a few more years before the technology pushes their appeal into the mainstream.

    Thank you for your 11th grade report. You may take your seat. Now let Jimmy do his report on why the economic downturn is bad.
    Reply
  • hellwig
    TheFace says it much more straighforward than I would have. This article just seems like a rehash of every SSD review on this site.

    SSD drives are not hot-swappable? That certainly is the #1 reason they aren't showing up in datacenters, if true.

    Enterprise is always slow to adopt, because they know better than to get screwed over with all the other early adopters. You always let a product mature before moving it into a critical environment. I had one of the first 16X CD-ROMs, but it was terrible, would lose track on all my CDs. Had to replace it with a good ole reliable 4X.

    Heck SCSI drives used to be so far behind in capacity it was embarrasing, but the reliability meant even though they were more expensive than regular IDE, they were worth it. Now we have a product thats more expensive, and less reliable, and people are surprised it isn't being bought up in mass quantities?
    Reply
  • Themurph
    TheFaceAlso, your supposed BIG NEWS is that the Enterprise market is slow to adopt something? OMG what a shocker!
    Lordy. I'm not going to get into an Internet shouting match, but let's identify the news in this article real quick:

    1) HP has piped in with an estimated date for SSD adoption
    2) HP has identified the key areas where SSDs need to improve before they can be considered for widespread enterprise use
    3) As a super-bonus, other research has identified two additional areas where SSD-related technology could benefit from improvement.

    I'm not going to lie to you: A few of the issues plaguing SSD enterprise adoption are the same ones that have either *been* there from the beginning, or are the very ones that everyone's talking about in the consumer market as well. I nevertheless thought that the hot-swap and controller points were of special interest, as those were two threads that I haven't been hearing in many of the ongoing SSD discussions lately.

    The enterprise is slow to adopt, as Hellwig points out. However, there's still a certain flurry of wide-eyed interest around SSDs that makes them seem like the savior of storage. These articles attempt to show why those in the industry agree and disagree, and more than that, what it will take to grow an SSD-friendly business environment.
    Reply
  • Kary
    I was somewhat confused by the drives being to slow (because of erasing to write) AND to fast for controllers to keep up with.

    Admittedly I haven't worked on THAT many databases, but aren't most of them read only anyway.

    I agree, this will be a niche market in the enterprise for awhile, but with costs coming down there should be more and more niches.
    Reply
  • blackened144
    It all depends on the application of the SSD as to whether it would be useful in a DC environment. We run supercomputer clusters at my place of business. Usually between 50 and 400 nodes. Data on these servers is only updated once a day, 5 days a week. At 100k writes, those drives would last us a long ass time. And since drive speed is usually the limiting factor of our clusters, SSD drives would be extremely nice. We are currently testing the FusionIO drive. It bypasses the SATA format and puts the SSD drive components directly on a PCIex card and it kicks some serious ass. Too bad they are so crazy expensive.
    Reply
  • blackened144
    I forgot to mention that the only reason we havent moved to SSD drives yet is because of the capacity. Went from 10k SCSI drives to 7200 RPM SAS drives because we needed more capacity and they dont offer spindle speeds faster than 7200RPM with 1TB drives. Hopefully that will change in the future.
    Reply
  • "Fixing Modern SSDs for the Enterprise"?

    OK, so which one of you broke them ??
    Reply