All of lore.kernel.org
 help / color / mirror / Atom feed
* [dmclock] reservation tag may be wrongly reduced
@ 2020-12-26  4:47 Xiangyang Yu
  0 siblings, 0 replies; only message in thread
From: Xiangyang Yu @ 2020-12-26  4:47 UTC (permalink / raw)
  To: ceph-devel

Hi cephers,
I have read the dmclock paper  and now I am reviewing the dmclock code.

When  a request is popped by a proportion tag ., then the reservation
tag need to reduceed. In my opinion, we only need to reduce by 1/r
in mclock or dmclock. (r is the  reservation inc )

In the ceph dmclock code, I find that reduced num is max (1, tag.rho),
I think it's wrong .  Code is showed as follows:

// data_mtx should be held when called
void reduce_reservation_tags(ImmediateTagCalc imm, ClientRec& client,
const RequestTag& tag) {
double res_offset =
client.info->reservation_inv * std::max(uint32_t(1), tag.rho);
for (auto& r : client.requests) {
r.tag.reservation -= res_offset;
}
}

If i have missed some information, please tell me.
thanks .

^ permalink raw reply	[flat|nested] only message in thread

only message in thread, other threads:[~2020-12-26  4:48 UTC | newest]

Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2020-12-26  4:47 [dmclock] reservation tag may be wrongly reduced Xiangyang Yu

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.