On 03/01/2018 12:23 AM, Michael Lyle wrote:

On 01/02/2018 12:53 AM, tang.junhui@zte.com.cn wrote:

quoted

If no front-end I/O coming, would this cause write-back IOs one by one
(one write-back IO issued must after the completion of the previous IO)?
though with zero delay time, is the write-back performance still good?

Hi Mike,

Yes-- this speeds up writeback during idle from 4k/sec to about 10MB/sec
on my slow 5400RPM disk.  (It writes back at about 20MB/sec with deep
queues in the same scenario).  So in this case it's about 2500x faster
than current idle writeback, but half as fast as writeback with full
device utilization.

Nice performance data!

With only one I/O outstanding at a time, if front-end I/O starts latency
for the first request is still good.

It is good idea, I like it.

Acked-by: Coly Li <redacted>

Thanks for the effort :-)

Coly Li