Ivan Betsis c818233b41 nvmf/rdma: Add WR batch rdma new transport option
With x86 and low queue depths with multiple QPs/initiators
there is a benefit from disable batch when we have randread IO pattern.

Testing environment: x86, Intel(R) Xeon(R) Silver 4116 CPU @ 2.10GH
Several results:
init_cores | QD | BS | Avg IOPs/BW without batch | Avg IOPS/BW with batch
  8 cores  | 4  | 4K |     1870087 / 7305        |     1594014 / 6226
  8 cores  | 8  | 4K |     1853573 / 7240        |     1576400 / 6157
  8 cores  | 16 | 4K |     1819643 / 7108        |     1569487 / 6130
  8 cores  | 32 | 4K |     1815467 / 7092        |     1569909 / 6132
 16 cores  | 4  | 4K |     1908018 / 7453        |     1566843 / 6120
 16 cores  | 8  | 4K |     1906081 / 7446        |     1562110 / 6102
 16 cores  | 16 | 4K |     1880706 / 7346        |     1555060 / 6074
 16 cores  | 32 | 4K |     1835878 / 7171        |     1548156 / 6046

Signed-off-by: Ivan Betsis <c_ivanb@mellanox.com>
Signed-off-by: Evgeniy Kochetov <evgeniik@mellanox.com>
Signed-off-by: Alexey Marchuk <alexeymar@mellanox.com>
Signed-off-by: Sasha Kotchubievsky <sashakot@mellanox.com>
Change-Id: Icdbbbdf83f137eb4f05bd2063268ee2a7d87335a
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/924
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Broadcom CI
2020-10-21 02:25:32 +00:00
..
2020-08-17 08:19:46 +00:00
2020-10-20 08:54:53 +00:00