common: Optimize the HDD mClock cost parameters.

With the earlier cost per io and cost per byte parameters, the
client ops performance was good but the recovery ops were taking
much longer to complete. The new cost parameters optimizes this
further so that the recovery ops progress faster while
maintaining similar client throughput.

This optimization is tested on both a small cluster and a larger
cluster in a scaled environment.

Fixes: https://tracker.ceph.com/issues/57529
Signed-off-by: Sridhar Seshasayee <sseshasa@redhat.com>
(cherry picked from commit 38655922870dd9dc36a23df6e01492da6ad44e1b)

diff --git a/src/common/options/osd.yaml.in b/src/common/options/osd.yaml.in
index 0911392fddcb4e483493bff37e132bbd825e6515..926c4bee6505a372b8a56bf6f7fa8ac5590841c6 100644 (file)
--- a/src/common/options/osd.yaml.in
+++ b/src/common/options/osd.yaml.in
@@ -973,7 +973,7 @@ options:
     cost factor in QoS calculations. Only considered for osd_op_queue = mclock_scheduler
   fmt_desc: Cost per IO in microseconds to consider per OSD (for rotational
     media)
-  default: 25000
+  default: 11400
   flags:
   - runtime
 - name: osd_mclock_cost_per_io_usec_ssd
@@ -1012,7 +1012,7 @@ options:
     = mclock_scheduler
   fmt_desc: Cost per byte in microseconds to consider per OSD (for rotational
     media)
-  default: 5.2
+  default: 2.6
   flags:
   - runtime
 - name: osd_mclock_cost_per_byte_usec_ssd