Answer to question #1
According to this article the ssd's seem to have quite good reliability and operational times before they blow up. To quote it:
Lifespan:
The longevity of SSDs, or lack of, as the case may be, is
blown way out of proportion. Most will probably be surprised to hear
NAND memory actually has a higher MTBF (Mean Time Before Failure) than
DRAM. How often does your DRAM fail once you've passed the 3 month
mark with it? Most SSDs have a MTBF of about 1 million hours plus
(it's actually 1 million writes). Has anyone actually done the math on
that? It works out to be over 20 yrs of continuous use; 24/7. This
assumes adequate "free" space. A "full" drive (using more than 85% of
its usable space) has very few (in the number of storage locations),
blocks/pages/cells to work with in its normal day to day operations.
This forces the SSD to use and reuse the same cells over and over
again. The algorithm used for wear leveling goes to hell when the
drive doesn't have enough free space for moving data. The cells that
comprise the free space end up being used over and over, and will fail
much sooner than those on the rest of the drive.
A lot (some say most) of the longevity of a drive actually has to do
with the amount of "over provisioning" on the drive. Over provisioning
is like spare parts for the drive (actually spare NAND). Artificial
numbers, but say you buy a 128GB SSD. That SSD may actually contain up
to 10% (12GB) of additional NAND that is not calculated into the drive
size stated by the manufacturer.
NAND memory can/and does go bad, it's a fact of NAND life. The cells
of NAND are little electronic traps, that trap electrons with "gate"
technology (although not exactly the same), just like the gates of a
transistor that runs your CPU. These "gates" over time can leak, be
susceptible to leakage from adjacent cells, or just plain fail, among
other things that render their use as problematic.
When an SSD's controller (firmware) determines that a cell is no
longer performing like it should, it will replace the data location
with one of the over provisioned blocks/pages, and no longer use the
"defective" location. Depending on the firmware's coding, this is
usually done on a "page" level (4 Kilobytes of space). It is basically
the same as when a magnetic HDD marks a sector of its spinning platter
as "bad"; although the HDD doesn't have the "spare parts" to replace
the bad sector.
[EDIT]: According to this AskUbuntu Question too, Ubuntu has support for TRIM since version 10.10 Maverick Meerkat, so this should help the drive stay healthy and do a good job prolonging its lifespan.
Answer to question #2
The SandyBridge processors and later have built in processor
extensions for handling encryption faster. This feature is called
AES instruction set. From wikipedia:
The purpose of the instruction set is to improve the speed of
applications performing encryption and decryption using the Advanced
Encryption Standard (AES).
And some instructions as an example:
- AESENC -> Perform one round of an AES encryption flow
- AESKEYGENASSIST -> Assist in AES key round generation
Answer to question #3
You have to understand that you do not encrypt the device itself, but the data it contains. The ssd is just another hard disk. Is the information contained encryption worthy? And more importantly, is it worth to have a performance hit everytime you want to do something? (be it negligible or not, it still will be there, and my guess is it will show when there are great amounts of data to be read or written)
