Commit 87b94e3b authored by Karol Latecki's avatar Karol Latecki Committed by Tomasz Zawadzki
Browse files

scripts/nvmf_perf: manually pick fio allowed cpus 



Fio "numa_cpu_nodes" parameter (which tells fio
which NUMA node to use to pick cpus for the job)
seemed promising, but manually adjusting
"cpus_allowed" seems to have better results.

Change-Id: I113566ad0dd1d32fb5a1e989725b257e3ccff4c0
Signed-off-by: default avatarKarol Latecki <karol.latecki@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/14842


Tested-by: default avatarSPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: default avatarKonrad Sztyber <konrad.sztyber@intel.com>
Reviewed-by: default avatarJim Harris <james.r.harris@intel.com>
parent d6db5988

scripts/perf/nvmf/run_nvmf.py

+14 −4
Original line number Diff line number Diff line
@@ -712,8 +712,9 @@ class Initiator(Server): 
                          "offset=0%",

                          "offset_increment=%s%%" % offset_inc])



    def gen_fio_numa_section(self, fio_filenames_list):

    def gen_fio_numa_section(self, fio_filenames_list, num_jobs):

        numa_stats = {}

        allowed_cpus = []

        for nvme in fio_filenames_list:

            nvme_numa = self.get_nvme_subsystem_numa(os.path.basename(nvme))

            numa_stats[nvme_numa] = numa_stats.setdefault(nvme_numa, 0) + 1
@@ -721,7 +722,14 @@ class Initiator(Server): 
        # Use the most common NUMA node for this chunk to allocate memory and CPUs

        section_local_numa = sorted(numa_stats.items(), key=lambda item: item[1], reverse=True)[0][0]



        return "\n".join(["numa_cpu_nodes=%s" % section_local_numa,

        for _ in range(num_jobs):

            try:

                allowed_cpus.append(str(self.available_cpus[section_local_numa].pop(0)))

            except IndexError:

                self.log.error("No more free CPU cores to use from allowed_cpus list!")

                raise



        return "\n".join(["cpus_allowed=%s" % ",".join(allowed_cpus),

                          "numa_mem_policy=prefer:%s" % section_local_numa])



    def gen_fio_config(self, rw, rwmixread, block_size, io_depth, subsys_no,
@@ -1287,6 +1295,7 @@ class KernelInitiator(Initiator): 
        return self.get_route_nic_numa(remote_nvme_ip.group(0))



    def gen_fio_filename_conf(self, subsystems, threads, io_depth, num_jobs=1, offset=False, offset_inc=0):

        self.available_cpus = self.get_numa_cpu_map()

        if len(threads) >= len(subsystems):

            threads = range(0, len(subsystems))
@@ -1320,7 +1329,7 @@ class KernelInitiator(Initiator): 
            if offset:

                offset_section = self.gen_fio_offset_section(offset_inc, num_jobs)



            numa_opts = self.gen_fio_numa_section(r)

            numa_opts = self.gen_fio_numa_section(r, num_jobs)



            filename_section = "\n".join([filename_section, header, disks, iodepth, numa_opts, offset_section, ""])
@@ -1408,6 +1417,7 @@ class SPDKInitiator(Initiator): 
        return json.dumps(bdev_cfg_section, indent=2)



    def gen_fio_filename_conf(self, subsystems, threads, io_depth, num_jobs=1, offset=False, offset_inc=0):

        self.available_cpus = self.get_numa_cpu_map()

        filename_section = ""

        if len(threads) >= len(subsystems):

            threads = range(0, len(subsystems))
@@ -1441,7 +1451,7 @@ class SPDKInitiator(Initiator): 
            if offset:

                offset_section = self.gen_fio_offset_section(offset_inc, num_jobs)



            numa_opts = self.gen_fio_numa_section(r)

            numa_opts = self.gen_fio_numa_section(r, num_jobs)



            filename_section = "\n".join([filename_section, header, disks, iodepth, numa_opts, offset_section, ""])