Improve instructions in README.md, remove error msg in scripts.

PlatformLab · May 24, 2018 · f7e31c0 · f7e31c0
1 parent fb9f774
commit f7e31c0
Show file tree

Hide file tree

Showing 3 changed files with 60 additions and 69 deletions.
diff --git a/README.md b/README.md
@@ -26,14 +26,12 @@ one of the memcached worker thread.
 
 ### How do I use memtier_skewsyn benchmark?
 
-1. Clone this repo and swith to the skewSynthetic branch.
+1. Clone this repo to your target directory ${MEMTIER_SKEWSYN_DIR}.
+The default branch is `skewSynthetic` branch.
 
     ```
-	MEMTIER_SKEWSYN_DIR=${HOME}/memtier_skewsyn
 	git clone https://github.com/PlatformLab/memtier_skewsyn.git ${MEMTIER_SKEWSYN_DIR}
 	cd ${MEMTIER_SKEWSYN_DIR}
-	git fetch
-	git checkout skewSynthetic
     ```
 
 2. Use the `scripts/prepare.sh` to install PerfUtils and compile memtier
@@ -49,7 +47,7 @@ By default, logs will be saved in `${MEMTIER_SKEWSYN_DIR}/exp_logs`
 
     1) Skew benchmark, only uses 1 client machine:
     ```
-    ./runSkew.sh <server> <key-min> <key-max> <data-size>  <iterations> <skew_bench> <log directory prefix: arachne/origin>
+    ./runSkew.sh <server> <key-min> <key-max> <data-size>  <iterations> <skew_bench> <log directory prefix>
     ```
     For example:
     ```
@@ -59,12 +57,41 @@ By default, logs will be saved in `${MEMTIER_SKEWSYN_DIR}/exp_logs`
     2) Colocation benchmark, uses 2 (or more) client machines (we suppose memtier_skewsyn is installed in
     the same path, or machines share the directory via nfs):
     ```
-    ./runSynthetic.sh <server> <key-min> <key-max> <data-size> <iterations> <synthetic_bench> <videos[0/1]> <prefix: arachne/origin> [list of clients...]
+    ./runSynthetic.sh <server> <key-min> <key-max> <data-size> <iterations> <synthetic_bench> <video[0/1]> <log directory prefix> [list of clients...]
     ```
     For example:
     ```
     ./runSynthetic.sh ${server} 1000000 9000000 200 1 workloads/Synthetic16.bench 0 arachne_0vid ${client1}
     ```
+    If you set `<video>` option to 0, it will run without background video processes.
+    If you are interested in the video colocation workload, please follow
+    instructions in the [following section](#how-to-play-with-video-colocation-workload).
+
+4. Parse logs:
+
+For skew benchmark, the logs loacte at
+`${MEMTIER_SKEWSYN_DIR}/exp_logs/<log directory prefix>_iters<iterations>_skew_logs/`
+directory.
+
+For colocation benchmark, the logs are at
+`${MEMTIER_SKEWSYN_DIR}/exp_logs/<log directory prefix>_iters<iterations>_synthetic_logs/`
+
+Inside the directory there are two directories: `latency/` and `throughput/`.
+`latency/` contains the latency log files on the main client machine.
+`throughput/` has throughput logs for both main client machine and ${client1}
+machine. The structures are as follows:
+
+```
+latency/ (skew workload does not record latencies)
+|---- latency_iter1.csv
+|---- latency_iter2.csv
+|...
+
+throughput/
+|---- qps_iter1.csv
+|---- qps_iter1_${client1}.csv (for colocation workload)
+|...
+```
 
 ### How to play with video colocation workload?
 
@@ -83,6 +110,8 @@ suppose the directory would be `${videoDir}`.
     in `${videoDir}/PerfUtils`, install nasm and x264 in `${videoDir}/install`.
     It will also automatically update the `$PATH` in `~/.bashrc`. By default, logs
     will be saved to `${videoDir}/exp_logs`.
+    This script can take a couple of minutes, especially the video downloading
+    part.
 
 2. Test video encoding alone by running:
     ```
@@ -104,6 +133,8 @@ modify `videoPATH` parameter in the `runSynthetic.sh` file. Then simply run:
     ```
     ./runSynthetic.sh ${server} 1000000 9000000 200 1 workloads/Synthetic16.bench 1 arachne_1vid ${client1}
     ```
-    You will find the corresponding log directories in `${videoDir}/exp_logs` and
-    `${MEMTIER_SKEWSYN_DIR}/exp_logs`.
+    You will find the corresponding log directories in
+    `${videoDir}/exp_logs/<log directory prefix>_iters<iterations>_synthetic_logs/` and
+    `${MEMTIER_SKEWSYN_DIR}/exp_logs/<log directory prefix>_iters<iterations>_synthetic_logs/`.
+    The log directory structures are the same as we described before.
 
diff --git a/scripts/runSkew.sh b/scripts/runSkew.sh
@@ -5,9 +5,8 @@
 scriptname=`basename "$0"`
 if [[ "$#" -lt 7 ]]; then
     echo "Usage: $scriptname <server> <key-min> <key-max> <data-size> " \
-         "<iterations> <skew_bench> <prefix: arachne/origin>" \
-         "[list of clients...]"
-    echo "Example: $scriptname n1 100000 200000 32 5 workloads/Skew.bench arachne n4..."
+         "<iterations> <skew_bench> <logDirPrefix>"
+    echo "Example: $scriptname n1 1000000 9000000 200 1 workloads/Skew16.bench arachne_test"
     exit
 fi
 
@@ -41,7 +40,7 @@ runlog=exp_logs/${prefix}_iters${iters}_skew_runlog.log
 echo "Saving logs to: $logdir"
 mkdir -p $logdir
 rm -rf $logdir/* # Clear previous logs
-rm $runlog
+rm -f $runlog
 
 # Load data into Memcached and warm up
 cmd="bash ${scriptPATH}/loadonly.sh $server $keymin $keymax $datasize $keyprefix"
@@ -53,43 +52,8 @@ read -p "Press enter to continue"
 # Execute experiments multiple times
 for iter in `seq 1 $iters`;
 do
-    # $9... will be client addresses, start clients!
-    if [[ "$#" -gt 7 ]]; then
-        # shift previous ones
-        for argn in `seq 1 7`;
-        do
-            shift
-        done
-
-        clientlogdir=$dirPATH/$logdir
-        clientbench=$dirPATH/$benchfile
-
-        for client in "$@";
-        do
-            echo "Starting client: $client ..."
-            clientqpsfile=${qpsprefix}_iter${iter}_${client}.csv
-            clientRunLog=$dirPATH/${runlog}_${client}
-            cmd="$dirPATH/memtier_benchmark -s $server \
-                 -p 11211 -P memcache_binary --clients $clients --threads $threads \
-                 --ratio $ratio --pipeline=$pipeline \
-                 --key-prefix=$keyprefix \
-                 --key-minimum=$keymin --key-maximum=$keymax \
-                 --data-size=$datasize --random-data --hide-histogram \
-                 --key-pattern=$keypattern --run-count=1 \
-                 --distinct-client-seed --randomize \
-                 --test-time=1 -b \
-                 --config-file=$clientbench --ir-dist=$irdist \
-                 --log-dir=${clientlogdir} \
-                 --log-qpsfile=$clientqpsfile"
-            #echo $cmd
-            sshcmd="ssh -p 22 $client \"nohup $cmd > $clientRunLog 2>&1 < /dev/null &\""
-            echo $sshcmd
-            ssh -p 22 $client "nohup $cmd > $clientRunLog 2>&1 < /dev/null & "
-        done
-    fi
 
     logqpsfile=${qpsprefix}_iter${iter}.csv
-    #latencyfile=${latencyprefix}_iter${iter}.csv
     cmd="./memtier_benchmark -s $server -p 11211 -P memcache_binary \
         --clients $clients --threads $threads --ratio $ratio \
         --pipeline=$pipeline \
@@ -101,11 +65,9 @@ do
         --config-file=$benchfile --ir-dist=$irdist --log-dir=$logdir \
         --log-qpsfile=$logqpsfile \
         --videos=$videos"
-        #--log-latencyfile=$latencyfile \
 
     # execute the commnad
     echo $cmd
-    # $cmd >> $runlog 2>&1
     $cmd 2>&1 | tee -a $runlog
 done
 
@@ -117,16 +79,16 @@ mkdir -p $qpsdir
 mkdir -p $latencydir
 
 # Clean those directories
-rm $qpsdir/*
-rm $latencydir/*
-mv $logdir/${qpsprefix}_iter* $qpsdir
-mv $logdir/${latencyprefix}_iter* $latencydir
+rm -f $qpsdir/*
+rm -f $latencydir/*
+mv $logdir/${qpsprefix}_iter* $qpsdir > /dev/null 2>&1
+mv $logdir/${latencyprefix}_iter* $latencydir > /dev/null 2>&1
 
 # Merge the data
-cmd="scripts/mergeStats.py ${qpsdir} ${prefix}_iters${iters}_qps"
-echo $cmd
-$cmd >> $runlog 2>&1
-
-cmd="scripts/mergeStats.py ${latencydir} ${prefix}_iters${iters}_latency"
-echo $cmd
-$cmd >> $runlog 2>&1
+#cmd="scripts/mergeStats.py ${qpsdir} ${prefix}_iters${iters}_qps"
+#echo $cmd
+#$cmd >> $runlog 2>&1
+#
+#cmd="scripts/mergeStats.py ${latencydir} ${prefix}_iters${iters}_latency"
+#echo $cmd
+#$cmd >> $runlog 2>&1
diff --git a/scripts/runSynthetic.sh b/scripts/runSynthetic.sh
@@ -7,7 +7,7 @@ if [[ "$#" -lt 8 ]]; then
     echo "Usage: $scriptname <server> <key-min> <key-max> <data-size>" \
          "<iterations> <synthetic_bench> <video[0/1]> <prefix: arachne/origin>" \
          "[list of clients...]"
-    echo "Example: $scriptname n1 100000 200000 32 5 workloads/LoadEstimator.bench 1 arachne n3 n4..."
+    echo "Example: $scriptname n1 1000000 9000000 200 1 workloads/Synthetic16.bench 0 arachne_0vid n3"
     exit
 fi
 
@@ -42,7 +42,7 @@ runlog=exp_logs/${prefix}_iters${iters}_synthetic_runlog.log
 echo "Saving logs to: $logdir"
 mkdir -p $logdir
 rm -rf $logdir/* # Clear previous logs
-rm $runlog
+rm -f $runlog
 
 # Load data into Memcached and warm up
 cmd="bash $scriptPATH/loadonly.sh $server $keymin $keymax $datasize $keyprefix"
@@ -95,10 +95,9 @@ do
         pipeline=1
         irdist="UNIFORM"
         benchfile=workloads/Synthetic16-master.bench
-        #benchfile=workloads/Synthetic16_1sec-master.bench
     fi
 
-    # Master usually run in non-blocking mode
+    # Start the main master node
     logqpsfile=${qpsprefix}_iter${iter}.csv
     latencyfile=${latencyprefix}_iter${iter}.csv
     cmd="./memtier_benchmark -s $server -p 11211 -P memcache_binary \
@@ -116,7 +115,6 @@ do
 
     # execute the commnad
     echo $cmd
-    # $cmd >> $runlog 2>&1
     $cmd 2>&1 | tee -a $runlog
 done
 
@@ -128,10 +126,10 @@ mkdir -p $qpsdir
 mkdir -p $latencydir
 
 # Clean those directories
-rm $qpsdir/*
-rm $latencydir/*
-mv $logdir/${qpsprefix}_iter* $qpsdir
-mv $logdir/${latencyprefix}_iter* $latencydir
+rm -f $qpsdir/*
+rm -f $latencydir/*
+mv $logdir/${qpsprefix}_iter* $qpsdir > /dev/null 2>&1
+mv $logdir/${latencyprefix}_iter* $latencydir > /dev/null 2>&1
 
 # Merge the data
 #cmd="scripts/mergeStats.py ${qpsdir} ${prefix}_iters${iters}_qps"