Self-healing flash for infinite life

Self-healing flash for infinite life

Summary: NAND flash is a wearing resource - like tires on a car - which creates many problems for high duty cycle storage. But a Taiwanese manufacturer demonstrates how to overcome that limit. Will it fly in the marketplace?

TOPICS: Storage

In an upcoming paper Macronix engineers show how flash cells can be rebuilt on the go.

Today's Multi Level Cell (MLC) NAND flash is typically specified to handle 10,000 writes - plenty for a phone or a point & shoot camera - but less than storage admins would like. If that limit could be raised - infinite writes would be nice! - flash storage would be more popular in data centers.

Heating flash memory cells has long been known to make them like new. But that's a lot of heat and power - impractical for the mobile devices that drive the NAND flash market.

But Macronix, a Taiwanese manufacturer, is demonstrating a
<blockquote> . . . novel self-healing flash, where a high temperature (>800°C), and short time annealing are generated by a built-in heater. </blockquote>

It takes just a few seconds of heating to return cells to a good state. By heating small groups of cells, they enable power-effective self-healing that over time could rebuild an entire flash chip when the device is connected to a power source.

The Storage Bits take
Macronix says that no commercial self-healing chip is imminent. And there's good economic reasons for that, despite the elegance of their design.

NAND flash markets are dominated by the ferocious cost pressures of high volume consumer products. About 95% of the flash market doesn't care about self-healing flash, so there's no way to justify the extra cost of their solution.

With well-engineered solutions, MLC flash can meet stringent enterprise requirements. The Macronix solution may become viable some day, but it isn't needed today.

Comments welcome, of course.

Does the limited life of MLC flash concern you and, if yes, why?

Topic: Storage

Kick off your day with ZDNet's daily email newsletter. It's the freshest tech news and opinion, served hot. Get it.


Log in or register to join the discussion
  • I didn't even have to read the whole article to know

    something like this would inevitably turn up as part of the sales pitch:

    " ... Macronix says that no commercial self-healing chip is imminent. And there's good economic reasons for that, despite the elegance of their design. "

    In other words, Macronix probably have gone well past proof of concept and have almost certainly got a secret prototype that has demonstrated all signs of vastly increased read/write capabilities ... but after talking with their CEO, Head of Accounting and Sales & Marketing people, have "magically" decided to not release the new technology.

    Now, Harris, you also state, " ... About 95% of the flash market doesn't care about self-healing flash, so there's no way to justify the extra cost of their solution." Well, why mention it if you don't even discuss "the extra cost"? I'd imagine in tough economic times the topic of 'Costs' is central & critical in a discussion like this. Why haven't Macronix (or you) provided production estimates in comparison to current NAND technology? A basic CBA would help too.

    This has all the traditional hallmarks of the old short-change, manufacturing policy: the self-healing technology is essentially a way to make SSD's last as long as - or potentially far longer than - traditional mechanical HDD's. So, heaven forbid, [manufacturer's name here] goes into predictable R&D overdrive to create a watered down version of [place technology's name here] ... and you guessed it ... [product name] wares out (a la deliberately designed for a short, product life span) thus forcing consumers (mock surprise!) .. into a product upgrade cycle.

    It's the same old, same old, self-perpetuating money making machine. I mean the formula works.

    So it's like this: MLC life span limitations concern me, to be sure - but nowhere near as much as an industry player (such as Macronix) having developed a technology that can greatly enhance data-center, server & workstation uptime - THEN deliberately withhold the technology from the public. Granted, Robin, i sound more than just a tad cynical. But that's nothing compared to the cynicism inferred via the actions of Macronix.

    I'm sure you're well aware by now that i don't like to fart around and exchange pretentious niceties. But then again, harsh economic times are hardly the time or place for pussy footing around.

    Apologies withheld.
  • This would make me switch

    This new technology would most likely make me switch my HDD for a SSD is my computers, but because of limited writes which gives them a shorter lifespan than regular HDDs I am still using the old mechanical drives.

    This would make me switch to enjoy the extra speed!

    Probably useless in smartphones and tablets because of their short lifespan, but could be useful as well in SD cards and USB drives, they would become more trust-able devices.
  • Nothing sinister IMHO

    NAND flash is a commodity business. Over 90% of the NAND market doesn't care about endurance. The remaining ≈10% does care, but they've already done a decent job of engineering around it.

    If Fusion-io or any other vendor of server SSDs came to Macronix with a large order, they'd build it. But getting that 1st design win won't be easy, and keeping up with the rate of cost improvement of generic MLC/TLC flash won't be easy either.

    Sure, it could happen. But I'm not expecting it and, it sounds like, neither is Macronix.

    R Harris
  • Dead CF module being used as an HDD

    We have a DNS server and since we don't reconfigure it, ever, we thought. "Given the lifetime of flash, and the fact we're not going to be writing to it much, if ever, let's use flash memory. No moving parts to wear out." So we set it up with noatime set using CF modules in adapters on the PATA connectors. (It's not a new server.) Of course, we had forgotten about /var/tmp. Last week, the DNS crashed. Post Mortem: The CF module has exceeded its write capacity for certain cells and ceased to work properly. (These weren't necessarily specifically DNS caching. /var/tmp holds a lot of things.)

    Would I purchase a self-healing flash module that would give me infinite writes? ABSOLUTELY!
    • mheartwood .. You have proved my point(s) exactly

      .. and yet, here Harris is saying:

      "... Sure, it could happen. But I'm not expecting it and, it sounds like, neither is Macronix."

      You and the (likely) the Admin's of tens of thousands of data centers / server farms, world-wide, that rely on at least 5 9's up operability would probably give their 'left ones' for self-healing flash module SSD's.

      But no, obviously the logic is with RH is to advertise the fact (a la "dangle a carrot"), then say, "Nope, you can't have it, nyah-nyah-nyah-nyah-nyah, NAH!"

      Thanks for the mixed messages, vagueries and inconsistencies Rob.

      The arrogance and audacity of it all.

      ... geez
  • in these bad economic times...

    Whilst the bad economic times make the manufacture of this new technology unappealing to the tiny minority - namely the manufacturer, it also makes this technology unavailable to the huge majority of people, namely the customers of this company (the end users), who could benefit dramatically from not having the financial burden of having to regularly replace the worn out flash. It is always the way with a capitalist society, where everything favors the tiny rich minority, and the majority suffer. Human beings - a waste of space.