Commit af0057e6 authored by Maciej Wawryk's avatar Maciej Wawryk Committed by Tomasz Zawadzki
Browse files

scripts/perf: Add effective QD counting 



During perf test we figured out, that global qd
settings is split between disks number in filename.
Setting this parameter (multiplied by section disks number
and numjobs) in filename section give as expected value.

Signed-off-by: default avatarMaciej Wawryk <maciejx.wawryk@intel.com>
Change-Id: I89a44e48f10da131e2e9128f3ff2fbfb197076ad
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/2749


Reviewed-by: default avatarKarol Latecki <karol.latecki@intel.com>
Reviewed-by: default avatarDarek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: default avatarTomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: default avatarShuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Tested-by: default avatarSPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
parent c0ff2954

scripts/perf/nvmf/README.md

+1 −1
Original line number Diff line number Diff line
@@ -104,7 +104,7 @@ other than -t, -s, -n and -a. 
Fio job parameters.



- bs: block size

- qd: io depth

- qd: io depth - Per connected fio filename target

- rw: workload mode

- rwmixread: percentage of reads in readwrite workloads

- run_time: time (in seconds) to run workload

scripts/perf/nvmf/run_nvmf.py

+11 −8
Original line number Diff line number Diff line
@@ -351,7 +351,6 @@ norandommap=1 
rw={rw}

rwmixread={rwmixread}

bs={block_size}

iodepth={io_depth}

time_based=1

ramp_time={ramp_time}

runtime={run_time}
@@ -388,13 +387,13 @@ runtime={run_time} 
            threads = range(0, len(subsystems))



        if "spdk" in self.mode:

            filename_section = self.gen_fio_filename_conf(subsystems, threads)

            filename_section = self.gen_fio_filename_conf(subsystems, threads, io_depth, num_jobs)

        else:

            filename_section = self.gen_fio_filename_conf(threads)

            filename_section = self.gen_fio_filename_conf(threads, io_depth, num_jobs)



        fio_config = fio_conf_template.format(ioengine=ioengine, spdk_conf=spdk_conf,

                                              rw=rw, rwmixread=rwmixread, block_size=block_size,

                                              io_depth=io_depth, ramp_time=ramp_time, run_time=run_time)

                                              ramp_time=ramp_time, run_time=run_time)

        if num_jobs:

            fio_config = fio_config + "numjobs=%s \n" % num_jobs

        if self.cpus_allowed is not None:
@@ -727,7 +726,7 @@ class KernelInitiator(Initiator): 
            self.remote_call("sudo %s disconnect -n %s" % (self.nvmecli_bin, subsystem[1]))

            time.sleep(1)



    def gen_fio_filename_conf(self, threads):

    def gen_fio_filename_conf(self, threads, io_depth, num_jobs=1):

        out, err = self.remote_call("lsblk -o NAME -nlp")

        nvme_list = [x for x in out.split("\n") if "nvme" in x]
@@ -747,7 +746,9 @@ class KernelInitiator(Initiator): 
        for i, r in enumerate(result):

            header = "[filename%s]" % i

            disks = "\n".join(["filename=/dev/%s" % x for x in r])

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

            job_section_qd = round((io_depth * len(r)) / num_jobs)

            iodepth = "iodepth=%s" % job_section_qd

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



        return filename_section
@@ -788,7 +789,7 @@ class SPDKInitiator(Initiator): 
        bdev_section = "\n".join([header, bdev_rows])

        return bdev_section



    def gen_fio_filename_conf(self, subsystems, threads):

    def gen_fio_filename_conf(self, subsystems, threads, io_depth, num_jobs=1):

        filename_section = ""

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

            threads = range(0, len(subsystems))
@@ -807,7 +808,9 @@ class SPDKInitiator(Initiator): 
        for i, r in enumerate(result):

            header = "[filename%s]" % i

            disks = "\n".join(["filename=%s" % x for x in r])

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

            job_section_qd = round((io_depth * len(r)) / num_jobs)

            iodepth = "iodepth=%s" % job_section_qd

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



        return filename_section