8c3ab23f4b
The lustre zpios-test simulates a reasonable lustre workload. It will create 128 threads, the same as a Lustre OSS, and then 4096 individual objects. Each objects is 16MiB in size and will be written/read in 1MiB from a random thread. This is fundamentally how we expect Lustre to behave for large IO intensive workloads.
67 lines
2.4 KiB
Bash
67 lines
2.4 KiB
Bash
#!/bin/bash
|
|
#
|
|
# Usage: zpios
|
|
# --threadcount -t =values
|
|
# --threadcount_low -l =value
|
|
# --threadcount_high -h =value
|
|
# --threadcount_incr -e =value
|
|
# --regioncount -n =values
|
|
# --regioncount_low -i =value
|
|
# --regioncount_high -j =value
|
|
# --regioncount_incr -k =value
|
|
# --offset -o =values
|
|
# --offset_low -m =value
|
|
# --offset_high -q =value
|
|
# --offset_incr -r =value
|
|
# --chunksize -c =values
|
|
# --chunksize_low -a =value
|
|
# --chunksize_high -b =value
|
|
# --chunksize_incr -g =value
|
|
# --regionsize -s =values
|
|
# --regionsize_low -A =value
|
|
# --regionsize_high -B =value
|
|
# --regionsize_incr -C =value
|
|
# --load -L =dmuio|ssf|fpp
|
|
# --pool -p =pool name
|
|
# --name -M =test name
|
|
# --cleanup -x
|
|
# --prerun -P =pre-command
|
|
# --postrun -R =post-command
|
|
# --log -G =log directory
|
|
# --regionnoise -I =shift
|
|
# --chunknoise -N =bytes
|
|
# --threaddelay -T =jiffies
|
|
# --verify -V
|
|
# --zerocopy -z
|
|
# --nowait -O
|
|
# --human-readable -H
|
|
# --verbose -v =increase verbosity
|
|
# --help -? =this help
|
|
|
|
ZPIOS_CMD="${ZPIOS} \
|
|
--load=dmuio,fpp \
|
|
--pool=${ZPOOL_NAME} \
|
|
--name=${ZPOOL_DESC} \
|
|
--threadcount=128 \
|
|
--regioncount=4096 \
|
|
--regionsize=16M \
|
|
--chunksize=1M \
|
|
--offset=0M \
|
|
--threaddelay=0 \
|
|
--cleanup \
|
|
--human-readable \
|
|
${ZPIOS_OPTIONS}"
|
|
|
|
zpios_start() {
|
|
if [ ${VERBOSE} ]; then
|
|
ZPIOS_CMD="${ZPIOS_CMD} --verbose"
|
|
echo ${ZPIOS_CMD}
|
|
fi
|
|
|
|
${ZPIOS_CMD} || exit 1
|
|
}
|
|
|
|
zpios_stop() {
|
|
[ ${VERBOSE} ] && echo
|
|
}
|