So you still need enough throughput on the traditional hardisks to
handle all the writes from memory if I understand correctly.
Ivar
matthew lagoe schreef:
To be frank if you get 2 years of use out of a drive, while that's not a
huge amount of time, its still quite good. Also in 2 years you will have
SSD's that you can replace it with for next to nothing that will be even
faster.
In regards to your 2nd question it is my understanding that the SLOG is
written to so that the OS can return a "write complete". ZFS only does a
write cycle once every 5 seconds, as such if you happened to lose power in
that 5 seconds you would lose all that data. the SLOG basically buffers the
writes to the SSD so in the event of power loss ect, the data is not lost
and is read from and written to the zpool on startup.
My numbers may not be exact above but the general idea is that rather then
writing to the slower drives, it buffers the writes to the fast SSD then
slowly writes to the zpool, making it so that especially small writes are
still fast, as the OS returns a "write complete" before its written to the
slow drives, while at the same time ensuring your data integrity with the
SLOG.
--
_______________________________________________
OpenIndiana-discuss mailing list
[email protected]
http://openindiana.org/mailman/listinfo/openindiana-discuss
