[AccessD] SSDs and BE storage

The truth be known Ken I pretty much agree with the general assessment of SSDs.  It is an immature 
technology, which you do have to study and plan for.  It is not a drop in replacement for rotating 
media.  It needs special handling.  Like any other technology it has a bathtub curve wherein you are 
going to have failures in the first 90 days, and if you get past that point...

And if you are in a position where you can afford SLC I highly recommend that.  SSD is definitely a 
technology where the die shrink works for density but against reliability.  OTOH it works against 
DRAM as well, yet I don't hear you railing against that.  In point of fact we are rapidly getting to 
the point where die shrink works against most technology reliability.

NASA does not send modern electronics into space precisely because high energy particles can cause 
significant charge displacement as the gates get so small, whereas the older technologies really did 
not have that issue.  Servers use ECC precisely for the same reason, the gates are getting so small 
that they have to have safeguards against errors that were previously unheard of.

So yes, I study the issues, I study the technology, I study the safeguards and I study the 
requirements in a way that I really don't need to do with rotating media.  But I need the benefits 
enough to make the effort worthwhile.

I used to swear by the top player and I used their product and I love the product that I bought from 
them two years ago.  Today I have switched to another player because my old choice is not reliable 
(due to using 28 nm chips).  I believe that long term, they will figure out how to use the 28 nm 
chips but today I won't use them, and there are players who agree and do not use them (yet).

It is my job to know this stuff if I am going to try and use it, to know what I can use it for and 
what I cannot.

John W. Colby
www.ColbyConsulting.com

On 9/22/2011 7:57 PM, Kenneth Ismert wrote:
> I'm sorry that you felt my remarks were unkind. You have done your own
> research, and have taken steps to mitigate the risks by limiting how your
> SSDs are used. And you are planning for failure with a robust backup scheme
> using traditional HDDs.
>
> Still, my first experience with an SSD out of the box was that the firmware
> had a critical bug, which required an update that wasn't available, because
> the maker needed a lot of time to test the revision. It seems likely that
> the real cause is that the drive, being a consumer level product, didn't
> have enough capacitance built-in to shut down gracefully during a power
> outage. Maybe we made a poor choice, but the device was from a dominant
> player in the industry. Not a confidence builder.
>
> Also, some of your own comments don't bolster your overall glowing
> endorsement. To wit: the SSD's with the 34 micron chips are good but the new
> 28 micron chips are bad? That sounds like caveat emptor to me. Problems with
> scaling down and firmware issues sound to me like indicators of a technology
> that is still not mature enough to be suitable as drop-in replacements for
> HDDs in a broad range of server applications.
>
> And yes, the one source I quoted does sound like a guy fishing for
> consulting work, but you conveniently neglected to criticize my other
> reference, who said things more central to my point, and who seems more
> credible.
>
> Certainly, the technology will mature, and adoption rates will climb.
> Operating systems that optimize for SSDs will become more popular. Right
> now, for our server installation, I'm content to wait it out.
>
> -Ken