[colug-432] SSD vs SD

Matt Simmons standalone.sysadmin at gmail.com
Sun Mar 17 14:00:35 EDT 2013


> Hmm.  Interesting.  So would a logical conclusion to this be that archival
> storage on an SSD is not recommended?  As in, for example, don't stick an
> ssd into an external storage bay, throw it in the closet and expect the data
> to be readable in a year?  Also, don't stick an ssd into a system that you
> only use (and then power down) a couple of times/month (or less frequently)?
> Basically - the SSD needs power.

Yes, that's correct. Essentially, it's a function of how "big" the
areas of doped silicon are. Larger cells (like those in SLC) will have
a longer shelf-life than smaller, or denser cells (thus cheaper) tend
to have shorter shelf lives.

For the record, though, it's really not good for spinning disks to sit
unused for years, either. The oil that lubricates the drive motors
tends to break down when it's not exercised for a while (some number
of years - I haven't found any solid numbers). There are archival
quality hard disks that feature long-life stable oils for the
lubrication.

There are several manufacturers who do bottom-tier archival storage
solutions where the drives sit in an array, powered off while not
being read from, and the array controllers periodically spin up the
disks to exercise the motors and check for failures. Even keeping a
spinning disk in an external enclosure and powering it up every
quarter or so should be enough to keep the drives alive. And even if
the drive DOES die, the data is still there, so low level forensics
can still recover it with almost certainty.

>
> It seems like by putting the OS on your SSD, you get a sort of narrow
> benefit - most of the libraries are loaded into memory on boot and obviously
> are never written to during the normal course of operations.  Most of your
> performance advantage is then at boot time, and you're back to the slower
> read/write times thereafter.  Obviously, I'm using mine for a pretty narrow
> case as well, but I've also got a very specific use case in mind.
>
> How are others taking advantage of the SSD performance?  I assume that most
> of us either have an SSD in our laptops or are mixing them with spinning
> disks in desktop systems.

Well, so for the OS itself, yes. You get boot times of a few seconds,
but once it's running, as you said, the OS loads from memory at the
same speed as ever. The REAL magic is when you load up your
applications. They open instantly. If you put %ProgramFiles% on solid
state media, then everything in it will also load fast (except for any
libraries living in system32). Also, I'm not that familiar with the
registry - I don't know if Windows loads the entire hive into memory
or if it has to parse it each time. If it loads from disk, that'll
effect the time, too.

Most people are loading their "moving parts" into SSD - that is, OS
and programs, with "finished" data moved onto spinning disks, as you
suggest. Capacities are growing and prices are coming down. I don't
think that spinning disks will be replaced for a good many years for
archival purposes, but for keeping data that's frequently accessed (or
that needs to do random IO), you just can't beat solid state.

> what is the right term to reference a spinning disk to differentiate it from an SSD?

Well, "spinning disk" is probably the most commonly used, although
"rotating media" or "disk drive" is used (since SSDs don't technically
have disks). I'm also fond of "spinning rust", which refers to the
bits of ferrous magnetic bits on the actual platter :-)


--Matt


-- 
LITTLE GIRL: But which cookie will you eat FIRST?
COOKIE MONSTER: Me think you have misconception of cookie-eating process.


More information about the colug-432 mailing list