I have following scenario:
- 450GiB cache device
- writeback enabled with target of 10%
What happens is when the space taken by writeback cache is less than 45GiB
then the bcache writes one block per second every second which isn't the best.
Example:
rate: 512/sec
dirty: 11.9G
target: 45.6G
proportional: -29.2M
derivative: 0
change: -29.2M/sec
next io: 328ms
This means that:
- one can hear drives clicking every second as heads snap into place
- disks never spin down even though box is fully idle
At this rate the writeback will be flushed in about a year which is
pointless and only uses more power as disks are being kept active even
tough they don't need to.
Ssometimes it reduces sequential read performance as disks are being
wrote in random spot every second.
Solution IMO would be either to schedule in interval longer than one
every second or let the the rate to go to 0 blocks per second when
there is not enough data in writeback cache.
Currently the writeback rate is capped to 1 block per second at minimum.
--
Best Regards
Jakub "Kubuxu" Sztandera
--
To unsubscribe from this list: send the line "unsubscribe linux-bcache" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
