On 8/4/2017 12:49 PM, Warner Losh via cctalk wrote:
On Fri, Aug 4, 2017 at 12:36 PM, Al Kossow via cctalk <cctalk@classiccmp.org
wrote:
On 8/4/17 11:14 AM, Warner Losh via cctalk wrote:
most SD cards can easily handle 100-200 writes
The issue would be things like the swap partition on a unix disk
or whatever the equivalent is under RSX
Right. But since Flash devices have a FTL that translates writes to new
locations in the NAND each time a logical block is written, there's no
issue here. This issue with swap hasn't been an issue with NAND flash since
early ~8MB CF cards (which is now almost 20 year old technology).
I have a lot of miles using CF and SD cards in embedded systems, using both
commercial grade and industrial parts since 2000 or so. I find it hard to
believe that RSX could generate 128GB of data enough times, even in a
swapping environment, to wear a card like that out. Even a more modest 8GB
would take a while to wear out under 100% write workload, which swapping
never is (since there's always readback for at least some of the pages
swapped out). Though I did base my computations on 1MB/s being the fastest
that Q-Bus can go, but that was my remembered performance from 3 decades
ago since I couldn't find an answer to that question with a quick google. I
shipped systems that were 100's if not 1000's times faster than the
pdp-11's that could generate much more data traffic to SD and CF cards, and
had very very few CF cards wear out. SD cards when we shipped needed to be
not the smallest capacity on the market to do well and even there only a
few cards wore out while I was doing this with them...
Warner
With everything @ 3.3 volts, you might as well use a ram dick cache
and back up dirty blocks on power fail, or power down, or reboot, as
a small battery would last forever, while main system is down.
