Thinkpads Forum

Okay .... will do, thanks

If you're going with an SSD primary and a large secondary drive, I'd recommend you use a program that will update an image file of your ssd onto your secondary drive at regular intervals. This way when your ssd fails (notice I said "when" not "if")*, you will have a current image of it to use on the replacement SSD.

*SSD drives, even without the premature failures and corrupt file problems that plague them, will all fail (by design). Unlike platter discs, an SSD has a finite number of disc writes, so with a platter, it was always a question of "if" it would fail, but with SSD, it's not a matter of "if", we know they will fail, and if the product was 100% reliable, we'd be able to know exactly when so we could simply replace it before it fails.

This doesn't mean it's a bad design, the number of times data can be written to an SSD is quite high, so you should get years of use out of them, but currently many have been failing prematurely.

It's also important to to have a firmware (or software) that logs all disc writes so that all sectors are used an equal number of times. Without this, you could have failed sectores appearing randomly, rendering a drive unreliable far before it's expected service life has expired.

If everything works perfect, no faulty firmware/hardware, and each sector is used an equal number of times, then you'll get a good long life from an SSD before it is fully consumed, but the fact remains that it is a consumable.

TuuS wrote:It's also important to to have a firmware (or software) that logs all disc writes so that all sectors are used an equal number of times. Without this, you could have failed sectores appearing randomly, rendering a drive unreliable far before it's expected service life has expired.

Are you referring to TRIM here or something else?

eecon wrote:Are you referring to TRIM here or something else?

TRIM is one solution, but I won't say it's the only possible solution. As the technology improves, we'll probably see better ways for the drive to control how often each sector is used.

I can't say for sure this is true, but I believe we are in a sort of "dark ages" of SSD right now. I think the technology is just emerging and is far from perfected. I suspect in the near future drives will have the ability to monitor their own health and usability, and will assure that no one sector is used beyond it's service life. In such a situation, you'd probably see the drives full capacity slightly decreasing with age, as sectors become flagged as unusable. It would also be nice if sectors reaching their "end of life" point could be flagged for data storage, so the user could elect to use them only for long term storage, perhaps for storing files as "read-only", as my understanding is you can read the sectors an infinite number of times, it's only writing that is limited and will cause the drive to fail at some point.

If this was possible, you could start out with a drive that is 100% read/write enabled, then a few years later the heavily used drive could be 80% read/write and 20% read/only, which would be far better then 80% read/write and 20% unusable.

One mindset that has to change is with current SSD technology, a harddrive has to be seen as a consumable. I don't think most consumers really understand that yet

Nimble Storage is doing some great things with cheap SSDs (and cheap SATA HDs) in their CS-series iSCSI SANs...
http://www.nimblestorage.com/blog/evalu ... e-systems/
http://www.nimblestorage.com/products/architecture/
These things REALLY smoke! Have got a handful of them in my client base already and we're all blown away... a big rack of FC 15k SAS drives still beats it and is still necessary for, say, a video production facility... but for normal business usage patterns, wow...
Obviously very different from how an SSD would be used in your laptop, but I imagine things could be held in a few GB of ram while the device figures out where is really best to put it...