A good place to start would be to do some math regarding the TBW of this model
and compare to your expected workload.
https://semiconductor.samsung.com/consumer-storage/support/warranty/?model=N0002506
Samsung's warranty only applies to drives that haven't been worked to death in
this way. I usually treat this TBW value as an easy EOL target for when the
drive may just stop due to excessive writes.
The 870 EVO 1TB has a warranty of 5 years, or 600 Terabytes Written, whichever
comes first. if you write 500GB worth of data every day that comes out to
182,500 GB (approximately 180TB) of data written in one year. in 5 years you
will have written 900TB of data. The warranty would end during the 3rd year.
This gives you a sense of how much data you actually need to write before you
hit that limit.
You probably aren't working the drive that hard.. so lets assume you will write
500GB every week, for 10 years.
gb_per_write = 500
writes_per_year = 52
tbw = (gb_per_write * writes_per_year) / 1024
print tbw
25.390625 TB written in 1 year
print tbw*10
253.90625 TB written in 10 years
What I don't know, is how many writes are associated with swap usage. Next step
for me would be to find out how much data is writing to my drive in an average
week, and caculate that in.
-Ben
--- Original Message ---
On Thursday, September 21st, 2023 at 2:00 AM, Keith Lofstrom
wrote:
> I appreciate the discussion (disagreements?) about file
> systems. Recently, I have been migrating all my systems
> from ancient "Redhat/CentOS" to Debian, and pretty much
> let Debian choose the file system and partitioning.
>
> Along with this, I am moving all systems to 1 TB Samsung
> 870 EVO 1TB SATA (except for my 12TB Dirvish backup drives).
> I don't store Big Stuff like music or videos, so even my
> largest drive, with files dating back 40 years, uses
> "only" 450 GB in / (including /bin and /usr and /home).
>
> -
>
> MY QUESTION isn't about file systems and partitioning,
> but about Samsung SSD Magician
> ... which allows me to set aside extra "userspace" SSD
> storage area for firmware error correction.
>
> Should I do so?
>
> Background info: SSD storage cells are fast but
> imperfect - many overwrites (thousands but not
> millions) will eventually wear out bit cells, in
> sectors that the drive firmware will replace with
> spare sectors ... "belonging" to disk firmware and
> hidden from user view.
>
> If the entire "visible" SSD is formatted, there will
> still be many hidden "firmware" spare sectors for
> error correction, but perhaps not enough if large
> areas of the disk are frequently overwritten ...
> ... which might happen with the swap partition on
> my ancient 3GB RAM laptops.
>
> My guess is that Linux will tend to use the "first
> sectors" of swap more often than the last sectors,
> so even if I make a much larger swap, the first
> sectors might wear out and fail. Perhaps Debian 12
> Bookworm is smart enough to spread usage throughout
> swap, rather than the first sectors.
>
> Again, I am VERY unlikely to use an entire terabyte
> drive, though I worry that Samsung's SSD Magician might
> do bad voodoo, especially with me mis-operating it.
>
> The 870 EVO 1TB drives are only $60, so I can afford
> to mess one up with an experiment, as long as I learn
> from the experience.
>
> Should I do so?
>
> Keith
> --
> Keith Lofstrom kei...@keithl.com
