Hi,
在 2023/06/13 22:07, Bart Van Assche 写道:
On 5/18/23 00:55, Yu Kuai wrote:
在 2023/05/18 10:23, Bart Van Assche 写道:
On 5/17/23 18:49, Yu Kuai wrote:
Currently, fair share from hctx_may_queue() requires two
atomic_read(active_queues and active_requests), I think this smoothing
method can be placed into get_tag fail path, for example, the more
times
a disk failed to get tag in a period of time, the more tag this disk
can
get, and all the information can be updated here(perhaps directly
record how many tags a disk can get, then hctx_may_queue() still only
require 2 atomic_read()).
That sounds interesting to me. Do you perhaps plan to implement this
approach and to post it as a patch?
Of course, I'll try to send a RFC patch.
Hi Kuai,
Has this RFC patch already been posted or did I perhaps miss it?
Sorry for the delay, I finished switch from not share to fair share
directly verion, however, I found that it's not that simple to add a
smoothing method, and I'm stuck here for now.
I'll try to send a RFC verion soon, the smoothing method may not be
flexible, but I'll make sure it can work for your simple case that
2 device share tags, and one only issue few io while the other issue
lots of io.
Thanks,
Kuai
Thanks,
Bart.
.
