ivampiresp/Spdk
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
ed486dc4ef
Spdk/test/nvme/perf/run_perf.sh

147 lines
8.5 KiB
Bash
Raw Normal View History

test/nvme: Test case 1 and 2 of NVME performance tests Measure Iops/core performance Change-Id: I5c2632b45c62cadbd46a67a7284ece9f77b70902 Signed-off-by: Pawel Niedzwiecki <pawelx.niedzwiecki@intel.com> Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/423170 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Reviewed-by: John Kariuki <John.K.Kariuki@intel.com> Reviewed-by: Karol Latecki <karol.latecki@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com> Reviewed-by: Ben Walker <benjamin.walker@intel.com>
2018-08-22 21:20:54 +00:00
#!/usr/bin/env bash
# Automated script that runs NVMe PMD/BDEV performance test.
# This script should be run as root. Please run the scripts/setup.sh before running this script to bind disks to VFIO/UIO driver
# This script takes the following parameters:
# "--run-time" - the run time for the workload in seconds
# "--ramp-time" - Fio will run the specified workload for this amount of time before logging any performance numbers
# "--cpu-allowed" - A comma-separated list of CPU cores used to run the workload. - When the spdk fio plugin is chosen, NVMe devices will
# be aligned to specific core according to their NUMA node. The script will try to align each core with devices matching core's
# on the same NUMA node first but if there are no devices left in the same NUMA node as the CPU Core then it will use devices on the other NUMA node.
# It is important to choose cores that will ensure best NUMA node allocation. For example, on a system with 8 devices on NUMA node
# 0 and 8 devices on NUMA node 1, cores 0-27 on numa node 0 and 28-55 on numa node 1, if test uses 16 disk and four cores
# then "--cpu-allowed=1,2,28,29" results in a NUMA-balanced configuration with 4 devices on each CPU core.
# However, if the test will use 10 CPU cores, then best option would be "--cpu-allowed=1,2,3,4,28,29,30,31,32,33" because cores 1-4 will be aligned with
# 2 devices on numa0 per core, cores 28-29 will be aligned with 2 devices on numa1 per core and cores 30-33 with 1 device on numa1 per core.
test/nvme: Add TC3 and TC4 to NVME performance test. Nvme latency tests IOPS vs. Latency at different queue depths Signed-off-by: Pawel Niedzwiecki <pawelx.niedzwiecki@intel.com> Change-Id: I1192a18fa05f36c74385d286e86db571d52b9224 Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/424063 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Reviewed-by: Karol Latecki <karol.latecki@intel.com> Reviewed-by: John Kariuki <John.K.Kariuki@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com>
2018-08-30 15:28:41 +00:00
# "--iodepth" - Number of I/Os to keep in flight per devices for SPDK fio_plugin and per job for kernel driver.
# "--driver" - "This parameter is used to set the ioengine and other fio parameters that determine how fio jobs issue I/O. SPDK supports two modes (nvme and bdev): to use the SPDK BDEV fio plugin set the value to bdev, set the value to nvme to use the SPDK NVME PMD.
# "There are 3 modes available for Linux Kernel driver: set the value to kernel-libaio to use the Linux asynchronous I/O engine,
# set the value to kernel-classic-polling to use the pvsynch2 ioengine in classic polling mode (100% load on the polling CPU core),
# set the value to kernel-hybrid-polling to use the pvsynch2 ioengine in hybrid polling mode where the polling thread sleeps for half the mean device execution time.
test/nvme: Test case 1 and 2 of NVME performance tests Measure Iops/core performance Change-Id: I5c2632b45c62cadbd46a67a7284ece9f77b70902 Signed-off-by: Pawel Niedzwiecki <pawelx.niedzwiecki@intel.com> Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/423170 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Reviewed-by: John Kariuki <John.K.Kariuki@intel.com> Reviewed-by: Karol Latecki <karol.latecki@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com> Reviewed-by: Ben Walker <benjamin.walker@intel.com>
2018-08-22 21:20:54 +00:00
# "--no-preconditioning" - skip preconditioning - Normally the script will precondition disks to put them in a steady state.
# However, preconditioning could be skipped, for example preconditiong has been already made and workload was 100% reads.
# "--disk-no" - use specified number of disks for test.
# "--repeat-no" Repeat each workolad specified number of times.
test/nvme: Add TC3 and TC4 to NVME performance test. Nvme latency tests IOPS vs. Latency at different queue depths Signed-off-by: Pawel Niedzwiecki <pawelx.niedzwiecki@intel.com> Change-Id: I1192a18fa05f36c74385d286e86db571d52b9224 Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/424063 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Reviewed-by: Karol Latecki <karol.latecki@intel.com> Reviewed-by: John Kariuki <John.K.Kariuki@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com>
2018-08-30 15:28:41 +00:00
# "--numjobs" - Number of fio threads running the workload.
# "--no-io-scaling" - Set number of iodepth to be per job instead per device for SPDK fio_plugin.
test/nvme: Test case 1 and 2 of NVME performance tests Measure Iops/core performance Change-Id: I5c2632b45c62cadbd46a67a7284ece9f77b70902 Signed-off-by: Pawel Niedzwiecki <pawelx.niedzwiecki@intel.com> Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/423170 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Reviewed-by: John Kariuki <John.K.Kariuki@intel.com> Reviewed-by: Karol Latecki <karol.latecki@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com> Reviewed-by: Ben Walker <benjamin.walker@intel.com>
2018-08-22 21:20:54 +00:00
# An Example Performance Test Run
# "./spdk/test/perf/run_perf.sh --run-time=600 --ramp-time=60 --cpu-allowed=28 --fio-bin=/usr/src/fio/fio\
# --rwmixread=100 --iodepth=256 --fio-plugin=bdev --no-preconditioning --disk-no=6"
# This command will run test using fio plugin for 600 seconds, 60 sec of ram time, randrw job with
# 100% reads with io depth 256 per disk, on 6 devices and skips preconditioning. Cpu core used for this test is
# core no 28.
BASE_DIR=$(readlink -f $(dirname $0))
. $BASE_DIR/common.sh
disk_names=$(get_disks $PLUGIN)
disks_numa=$(get_numa_node $PLUGIN "$disk_names")
cores=$(get_cores "$CPUS_ALLOWED")
no_cores=($cores)
no_cores=${#no_cores[@]}
if $PRECONDITIONING; then
test/nvme: Add TC3 and TC4 to NVME performance test. Nvme latency tests IOPS vs. Latency at different queue depths Signed-off-by: Pawel Niedzwiecki <pawelx.niedzwiecki@intel.com> Change-Id: I1192a18fa05f36c74385d286e86db571d52b9224 Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/424063 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Reviewed-by: Karol Latecki <karol.latecki@intel.com> Reviewed-by: John Kariuki <John.K.Kariuki@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com>
2018-08-30 15:28:41 +00:00
HUGEMEM=8192 $ROOT_DIR/scripts/setup.sh
test/nvme: Test case 1 and 2 of NVME performance tests Measure Iops/core performance Change-Id: I5c2632b45c62cadbd46a67a7284ece9f77b70902 Signed-off-by: Pawel Niedzwiecki <pawelx.niedzwiecki@intel.com> Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/423170 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Reviewed-by: John Kariuki <John.K.Kariuki@intel.com> Reviewed-by: Karol Latecki <karol.latecki@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com> Reviewed-by: Ben Walker <benjamin.walker@intel.com>
2018-08-22 21:20:54 +00:00
cp $BASE_DIR/config.fio.tmp $BASE_DIR/config.fio
preconditioning
rm -f $BASE_DIR/config.fio
fi
test/nvme: Add TC3 and TC4 to NVME performance test. Nvme latency tests IOPS vs. Latency at different queue depths Signed-off-by: Pawel Niedzwiecki <pawelx.niedzwiecki@intel.com> Change-Id: I1192a18fa05f36c74385d286e86db571d52b9224 Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/424063 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Reviewed-by: Karol Latecki <karol.latecki@intel.com> Reviewed-by: John Kariuki <John.K.Kariuki@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com>
2018-08-30 15:28:41 +00:00
#Kernel Classic Polling ioengine parameters
if [ $PLUGIN = "kernel-classic-polling" ]; then
$ROOT_DIR/scripts/setup.sh reset
fio_ioengine_opt="--ioengine=pvsync2 --hipri=100"
for disk in $disk_names; do
echo -1 > /sys/block/$disk/queue/io_poll_delay
done
#Kernel Hybrid Polling ioengine parameter
elif [ $PLUGIN = "kernel-hybrid-polling" ]; then
$ROOT_DIR/scripts/setup.sh reset
fio_ioengine_opt="--ioengine=pvsync2 --hipri=100"
for disk in $disk_names; do
echo 0 > /sys/block/$disk/queue/io_poll_delay
done
elif [ $PLUGIN = "kernel-libaio" ]; then
$ROOT_DIR/scripts/setup.sh reset
fio_ioengine_opt="--ioengine=libaio"
fi
test/nvme: Test case 1 and 2 of NVME performance tests Measure Iops/core performance Change-Id: I5c2632b45c62cadbd46a67a7284ece9f77b70902 Signed-off-by: Pawel Niedzwiecki <pawelx.niedzwiecki@intel.com> Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/423170 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Reviewed-by: John Kariuki <John.K.Kariuki@intel.com> Reviewed-by: Karol Latecki <karol.latecki@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com> Reviewed-by: Ben Walker <benjamin.walker@intel.com>
2018-08-22 21:20:54 +00:00
result_dir=perf_results_${BLK_SIZE}BS_${IODEPTH}QD_${RW}_${MIX}MIX_${PLUGIN}_${date}
test/nvme: Add TC3 and TC4 to NVME performance test. Nvme latency tests IOPS vs. Latency at different queue depths Signed-off-by: Pawel Niedzwiecki <pawelx.niedzwiecki@intel.com> Change-Id: I1192a18fa05f36c74385d286e86db571d52b9224 Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/424063 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Reviewed-by: Karol Latecki <karol.latecki@intel.com> Reviewed-by: John Kariuki <John.K.Kariuki@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com>
2018-08-30 15:28:41 +00:00
mkdir -p $BASE_DIR/results/$result_dir
test/nvme: Test case 1 and 2 of NVME performance tests Measure Iops/core performance Change-Id: I5c2632b45c62cadbd46a67a7284ece9f77b70902 Signed-off-by: Pawel Niedzwiecki <pawelx.niedzwiecki@intel.com> Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/423170 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Reviewed-by: John Kariuki <John.K.Kariuki@intel.com> Reviewed-by: Karol Latecki <karol.latecki@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com> Reviewed-by: Ben Walker <benjamin.walker@intel.com>
2018-08-22 21:20:54 +00:00
result_file=$BASE_DIR/results/$result_dir/perf_results_${BLK_SIZE}BS_${IODEPTH}QD_${RW}_${MIX}MIX_${PLUGIN}_${date}.csv
test/nvme: Add TC3 and TC4 to NVME performance test. Nvme latency tests IOPS vs. Latency at different queue depths Signed-off-by: Pawel Niedzwiecki <pawelx.niedzwiecki@intel.com> Change-Id: I1192a18fa05f36c74385d286e86db571d52b9224 Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/424063 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Reviewed-by: Karol Latecki <karol.latecki@intel.com> Reviewed-by: John Kariuki <John.K.Kariuki@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com>
2018-08-30 15:28:41 +00:00
unset iops_disks bw mean_lat_disks_usec p99_lat_disks_usec p99_99_lat_disks_usec stdev_disks_usec
test/nvme: Test case 1 and 2 of NVME performance tests Measure Iops/core performance Change-Id: I5c2632b45c62cadbd46a67a7284ece9f77b70902 Signed-off-by: Pawel Niedzwiecki <pawelx.niedzwiecki@intel.com> Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/423170 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Reviewed-by: John Kariuki <John.K.Kariuki@intel.com> Reviewed-by: Karol Latecki <karol.latecki@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com> Reviewed-by: Ben Walker <benjamin.walker@intel.com>
2018-08-22 21:20:54 +00:00
echo "run-time,ramp-time,fio-plugin,QD,block-size,num-cpu-cores,workload,workload-mix" > $result_file
printf "%s,%s,%s,%s,%s,%s,%s,%s\n" $RUNTIME $RAMP_TIME $PLUGIN $IODEPTH $BLK_SIZE $no_cores $RW $MIX >> $result_file
test/nvme: Add TC3 and TC4 to NVME performance test. Nvme latency tests IOPS vs. Latency at different queue depths Signed-off-by: Pawel Niedzwiecki <pawelx.niedzwiecki@intel.com> Change-Id: I1192a18fa05f36c74385d286e86db571d52b9224 Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/424063 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Reviewed-by: Karol Latecki <karol.latecki@intel.com> Reviewed-by: John Kariuki <John.K.Kariuki@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com>
2018-08-30 15:28:41 +00:00
echo "num_of_disks,iops,avg_lat[usec],p99[usec],p99.99[usec],stdev[usec],avg_slat[usec],avg_clat[usec],bw[Kib/s]" >> $result_file
test/nvme: Test case 1 and 2 of NVME performance tests Measure Iops/core performance Change-Id: I5c2632b45c62cadbd46a67a7284ece9f77b70902 Signed-off-by: Pawel Niedzwiecki <pawelx.niedzwiecki@intel.com> Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/423170 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Reviewed-by: John Kariuki <John.K.Kariuki@intel.com> Reviewed-by: Karol Latecki <karol.latecki@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com> Reviewed-by: Ben Walker <benjamin.walker@intel.com>
2018-08-22 21:20:54 +00:00
#Run each workolad $REPEAT_NO times
for (( j=0; j < $REPEAT_NO; j++ ))
do
#Start with $DISKNO disks and remove 2 disks for each run to avoid preconditioning before each run.
for (( k=$DISKNO; k >= 1; k-=2 ))
do
cp $BASE_DIR/config.fio.tmp $BASE_DIR/config.fio
echo "" >> $BASE_DIR/config.fio
#The SPDK fio plugin supports submitting/completing I/Os to multiple SSDs from a single thread.
#Therefore, the per thread queue depth is set to the desired IODEPTH/device X the number of devices per thread.
test/nvme: Add TC3 and TC4 to NVME performance test. Nvme latency tests IOPS vs. Latency at different queue depths Signed-off-by: Pawel Niedzwiecki <pawelx.niedzwiecki@intel.com> Change-Id: I1192a18fa05f36c74385d286e86db571d52b9224 Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/424063 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Reviewed-by: Karol Latecki <karol.latecki@intel.com> Reviewed-by: John Kariuki <John.K.Kariuki@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com>
2018-08-30 15:28:41 +00:00
if [ "$PLUGIN" = "nvme" ] || [ "$PLUGIN" = "bdev" ] && [ "$NOIOSCALING" = false ]; then
qd=$(( $IODEPTH * $k ))
else
qd=$IODEPTH
fi
create_fio_config $k $PLUGIN "$disk_names" "$disks_numa" "$cores"
desc="Running Test: Blocksize=${BLK_SIZE} Workload=$RW MIX=${MIX} qd=${IODEPTH} io_plugin/driver=$PLUGIN"
if [ $PLUGIN = "nvme" ] || [ $PLUGIN = "bdev" ]; then
run_spdk_nvme_fio $PLUGIN "--runtime=$RUNTIME" "--ramp_time=$RAMP_TIME" "--bs=$BLK_SIZE"\
"--rw=$RW" "--rwmixread=$MIX" "--iodepth=$qd" "--output=$NVME_FIO_RESULTS" "--time_based=1"\
"--numjobs=$NUMJOBS" "--description=$desc" "-log_avg_msec=250"\
"--write_lat_log=$BASE_DIR/results/$result_dir/perf_lat_$${BLK_SIZE}BS_${IODEPTH}QD_${RW}_${MIX}MIX_${PLUGIN}_${date}_${k}disks_${j}"
else
run_nvme_fio $fio_ioengine_opt "--runtime=$RUNTIME" "--ramp_time=$RAMP_TIME" "--bs=$BLK_SIZE"\
"--rw=$RW" "--rwmixread=$MIX" "--iodepth=$qd" "--output=$NVME_FIO_RESULTS" "--time_based=1"\
"--numjobs=$NUMJOBS" "--description=$desc" "-log_avg_msec=250"\
"--write_lat_log=$BASE_DIR/results/$result_dir/perf_lat_${BLK_SIZE}BS_${IODEPTH}QD_${RW}_${MIX}MIX_${PLUGIN}_${date}_${k}disks_${j}"
fi
test/nvme: Test case 1 and 2 of NVME performance tests Measure Iops/core performance Change-Id: I5c2632b45c62cadbd46a67a7284ece9f77b70902 Signed-off-by: Pawel Niedzwiecki <pawelx.niedzwiecki@intel.com> Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/423170 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Reviewed-by: John Kariuki <John.K.Kariuki@intel.com> Reviewed-by: Karol Latecki <karol.latecki@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com> Reviewed-by: Ben Walker <benjamin.walker@intel.com>
2018-08-22 21:20:54 +00:00
#Store values for every number of used disks
iops_disks[$k]=$((${iops_disks[$k]} + $(get_results iops $MIX)))
mean_lat_disks_usec[$k]=$((${mean_lat_disks_usec[$k]} + $(get_results mean_lat_usec $MIX)))
p99_lat_disks_usec[$k]=$((${p99_lat_disks_usec[$k]} + $(get_results p99_lat_usec $MIX)))
p99_99_lat_disks_usec[$k]=$((${p99_99_lat_disks_usec[$k]} + $(get_results p99_99_lat_usec $MIX)))
stdev_disks_usec[$k]=$((${stdev_disks_usec[$k]} + $(get_results stdev_usec $MIX)))
test/nvme: Add TC3 and TC4 to NVME performance test. Nvme latency tests IOPS vs. Latency at different queue depths Signed-off-by: Pawel Niedzwiecki <pawelx.niedzwiecki@intel.com> Change-Id: I1192a18fa05f36c74385d286e86db571d52b9224 Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/424063 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Reviewed-by: Karol Latecki <karol.latecki@intel.com> Reviewed-by: John Kariuki <John.K.Kariuki@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com>
2018-08-30 15:28:41 +00:00
mean_slat_disks_usec[$k]=$((${mean_slat_disks_usec[$k]} + $(get_results mean_slat_usec $MIX)))
mean_clat_disks_usec[$k]=$((${mean_clat_disks_usec[$k]} + $(get_results mean_clat_usec $MIX)))
bw[$k]=$((${bw[$k]} + $(get_results bw_Kibs $MIX)))
test/nvme: Test case 1 and 2 of NVME performance tests Measure Iops/core performance Change-Id: I5c2632b45c62cadbd46a67a7284ece9f77b70902 Signed-off-by: Pawel Niedzwiecki <pawelx.niedzwiecki@intel.com> Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/423170 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Reviewed-by: John Kariuki <John.K.Kariuki@intel.com> Reviewed-by: Karol Latecki <karol.latecki@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com> Reviewed-by: Ben Walker <benjamin.walker@intel.com>
2018-08-22 21:20:54 +00:00
cp $NVME_FIO_RESULTS $BASE_DIR/results/$result_dir/perf_results_${MIX}_${PLUGIN}_${no_cores}cpus_${date}_${k}_disks_${j}.json
cp $BASE_DIR/config.fio $BASE_DIR/results/$result_dir/config_${MIX}_${PLUGIN}_${no_cores}cpus_${date}_${k}_disks_${j}.fio
rm -f $BASE_DIR/config.fio
#if tested on only one number of disk
if $ONEWORKLOAD; then
break
fi
done
done
#Write results to csv file
for (( k=$DISKNO; k >= 1; k-=2 ))
do
test/nvme: Add TC3 and TC4 to NVME performance test. Nvme latency tests IOPS vs. Latency at different queue depths Signed-off-by: Pawel Niedzwiecki <pawelx.niedzwiecki@intel.com> Change-Id: I1192a18fa05f36c74385d286e86db571d52b9224 Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/424063 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Reviewed-by: Karol Latecki <karol.latecki@intel.com> Reviewed-by: John Kariuki <John.K.Kariuki@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com>
2018-08-30 15:28:41 +00:00
iops_disks[$k]=$((${iops_disks[$k]} / $REPEAT_NO))
mean_lat_disks_usec[$k]=$((${mean_lat_disks_usec[$k]} / $REPEAT_NO))
p99_lat_disks_usec[$k]=$((${p99_lat_disks_usec[$k]} / $REPEAT_NO))
p99_99_lat_disks_usec[$k]=$((${p99_99_lat_disks_usec[$k]} / $REPEAT_NO))
stdev_disks_usec[$k]=$((${stdev_disks_usec[$k]} / $REPEAT_NO))
mean_slat_disks_usec[$k]=$((${mean_slat_disks_usec[$k]} / $REPEAT_NO))
mean_clat_disks_usec[$k]=$((${mean_clat_disks_usec[$k]} / $REPEAT_NO))
bw[$k]=$((${bw[$k]} / $REPEAT_NO))
test/nvme: Test case 1 and 2 of NVME performance tests Measure Iops/core performance Change-Id: I5c2632b45c62cadbd46a67a7284ece9f77b70902 Signed-off-by: Pawel Niedzwiecki <pawelx.niedzwiecki@intel.com> Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/423170 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Reviewed-by: John Kariuki <John.K.Kariuki@intel.com> Reviewed-by: Karol Latecki <karol.latecki@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com> Reviewed-by: Ben Walker <benjamin.walker@intel.com>
2018-08-22 21:20:54 +00:00
test/nvme: Add TC3 and TC4 to NVME performance test. Nvme latency tests IOPS vs. Latency at different queue depths Signed-off-by: Pawel Niedzwiecki <pawelx.niedzwiecki@intel.com> Change-Id: I1192a18fa05f36c74385d286e86db571d52b9224 Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/424063 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Reviewed-by: Karol Latecki <karol.latecki@intel.com> Reviewed-by: John Kariuki <John.K.Kariuki@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com>
2018-08-30 15:28:41 +00:00
printf "%s,%s,%s,%s,%s,%s,%s,%s,%s\n" ${k} ${iops_disks[$k]} ${mean_lat_disks_usec[$k]} ${p99_lat_disks_usec[$k]}\
${p99_99_lat_disks_usec[$k]} ${stdev_disks_usec[$k]} ${mean_slat_disks_usec[$k]} ${mean_clat_disks_usec[$k]} ${bw[$k]} >> $result_file
test/nvme: Test case 1 and 2 of NVME performance tests Measure Iops/core performance Change-Id: I5c2632b45c62cadbd46a67a7284ece9f77b70902 Signed-off-by: Pawel Niedzwiecki <pawelx.niedzwiecki@intel.com> Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/423170 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Reviewed-by: John Kariuki <John.K.Kariuki@intel.com> Reviewed-by: Karol Latecki <karol.latecki@intel.com> Reviewed-by: Jim Harris <james.r.harris@intel.com> Reviewed-by: Ben Walker <benjamin.walker@intel.com>
2018-08-22 21:20:54 +00:00
#if tested on only one numeber of disk
if $ONEWORKLOAD; then
break
fi
done
rm -f $BASE_DIR/bdev.conf $BASE_DIR/config.fio
Reference in New Issue Copy Permalink