- ptlrpc_ping_interpret is needless:

[fs/lustre-release.git] / lustre-iokit / sgpdd-survey / sgpdd-survey

diff --git a/lustre-iokit/sgpdd-survey/sgpdd-survey b/lustre-iokit/sgpdd-survey/sgpdd-survey
index 8f7cd1f..52207f1 100755 (executable)
--- a/lustre-iokit/sgpdd-survey/sgpdd-survey
+++ b/lustre-iokit/sgpdd-survey/sgpdd-survey
@@ -6,9 +6,9 @@
 # the SCSI devices to measure
 scsidevs="/dev/sde /dev/sdh"
 
 # the SCSI devices to measure
 scsidevs="/dev/sde /dev/sdh"
 

-# result file prefix
+# result file prefix.  date/time+hostname makes unique

 # NB ensure the path exists if it includes subdirs
 # NB ensure the path exists if it includes subdirs

-rslt=/tmp/sg_dd_rslt
+rslt=/tmp/sgpdd_survey_`date +%F@%R`_`uname -n`

 
 # what to do (read or write)
 actions="write read"
 
 # what to do (read or write)
 actions="write read"


@@ -23,21 +23,22 @@ rszhi=1024
 
 # Concurrent regions per device
 crglo=1
 
 # Concurrent regions per device
 crglo=1

-crghi=128
+crghi=256

 
 # threads to share between concurrent regions per device
 
 # threads to share between concurrent regions per device

-# NB survey skips over #thr < #regions
+# NB survey skips over #thr < #regions and #thr/#regions > SG_MAX_QUEUE

 thrlo=1
 thrlo=1

-thrhi=128
+thrhi=4096

 
 #####################################################################
 
 #####################################################################

+# leave the rest of this alone unless you know what you're doing...

 
 

-# disk block size (Bytes)
+# sgp_dd's idea of disk sector size (Bytes)

 bs=512
 bs=512

-
-# max # threads per individual sgp_dd instance
+# and max # threads one instance will spawn

 SG_MAX_QUEUE=16
 
 SG_MAX_QUEUE=16
 

+# map given device names into SG device names

 i=0
 devs=()
 for d in $scsidevs; do
 i=0
 devs=()
 for d in $scsidevs; do


@@ -50,12 +51,10 @@ for d in $scsidevs; do
 done
 ndevs=${#devs[@]}
 
 done
 ndevs=${#devs[@]}
 

-start=`date +%F@%R`
-rsltf=${rslt}_${start}.summary
+rsltf=${rslt}.summary
+workf=${rslt}.detail

 echo -n > $rsltf
 echo -n > $rsltf

-workf=${rslt}_${start}.detail

 echo -n > $workf
 echo -n > $workf

-tmpf=${rslt}_${start}.tmp

 
 print_summary () {
     if [ "$1" = "-n" ]; then
 
 print_summary () {
     if [ "$1" = "-n" ]; then


@@ -80,7 +79,7 @@ for ((rsz=$rszlo;rsz<=$rszhi;rsz*=2)); do
            # show computed parameters
            actual_rsz=$((bpt*bs/1024))
            actual_size=$((bs*count*crg/1024))
            # show computed parameters
            actual_rsz=$((bpt*bs/1024))
            actual_size=$((bs*count*crg/1024))

-           str=`printf 'total_size %8dK rsz %4d crg %5d thr %3d ' \
+           str=`printf 'total_size %8dK rsz %4d crg %5d thr %5d ' \

                         $((actual_size*ndevs)) $actual_rsz $((crg*ndevs)) $((thr*ndevs))`
            echo "==============> $str" >> $workf
            print_summary -n "$str"
                         $((actual_size*ndevs)) $actual_rsz $((crg*ndevs)) $((thr*ndevs))`
            echo "==============> $str" >> $workf
            print_summary -n "$str"


@@ -89,9 +88,11 @@ for ((rsz=$rszlo;rsz<=$rszhi;rsz*=2)); do
                print_summary "ENOMEM"
                continue
            fi
                print_summary "ENOMEM"
                continue
            fi

+           # run tests

            for action in $actions; do
                print_summary -n "$action "
                echo "=====> $action" >> $workf
            for action in $actions; do
                print_summary -n "$action "
                echo "=====> $action" >> $workf

+               tmpf=${workf}_tmp

                 # start test
                t0=`date +%s.%N`
                for ((i=0;i<ndevs;i++)); do
                 # start test
                t0=`date +%s.%N`
                for ((i=0;i<ndevs;i++)); do


@@ -113,7 +114,7 @@ for ((rsz=$rszlo;rsz<=$rszhi;rsz*=2)); do
                done 
                wait
                t1=`date +%s.%N`
                done 
                wait
                t1=`date +%s.%N`

-               # collect all results in 1 file
+               # collect/check individual stats

                echo > $tmpf
                ok=0
                for ((i=0;i<ndevs;i++)); do
                echo > $tmpf
                ok=0
                for ((i=0;i<ndevs;i++)); do


@@ -131,11 +132,11 @@ for ((rsz=$rszlo;rsz<=$rszhi;rsz*=2)); do
                    print_summary -n "$((ndevs*crg - ok)) failed "
                else
                    # compute MB/sec from elapsed
                    print_summary -n "$((ndevs*crg - ok)) failed "
                else
                    # compute MB/sec from elapsed

-                   bw=`awk "BEGIN {printf \"%7.2f MB/s\", $actual_size / (( $t1 - $t0 ) * 1024); exit}"`
+                   bw=`awk "BEGIN {printf \"%7.2f MB/s\", $actual_size * $ndevs / (( $t1 - $t0 ) * 1024); exit}"`

                    # compute MB/sec from nregions*slowest
                    check=`awk < $tmpf \
                        '/time to transfer data/ {mb=$8/1.048576; if (n == 0 || mb < min) min = mb; n++}\
                    # compute MB/sec from nregions*slowest
                    check=`awk < $tmpf \
                        '/time to transfer data/ {mb=$8/1.048576; if (n == 0 || mb < min) min = mb; n++}\

-                       END {printf "%3d x %6.2f = %7.2f MB/s", n, min, min * n}'`
+                       END {printf "%5d x %6.2f = %7.2f MB/s", n, min, min * n}'`

                    print_summary -n "$bw $check "
                fi
                rm $tmpf
                    print_summary -n "$bw $check "
                fi
                rm $tmpf