Adding tools for running experiments

store/tools/key_generator.py

+import random
+import string
+import sys
+
+charset = string.ascii_uppercase + string.ascii_lowercase + string.digits
+
+for i in range(int(sys.argv[1])):
+  rkey = "".join(random.choice(charset) for j in range(64))
+  print rkey

store/tools/process_logs.py

+import sys
+
+start, end = -1.0, -1.0
+
+duration = float(sys.argv[2])
+warmup = duration/3.0
+
+tLatency = []
+sLatency = []
+fLatency = []
+
+tExtra = 0.0
+sExtra = 0.0
+fExtra = 0.0
+
+xLatency = []
+
+for line in open(sys.argv[1]):
+  if line.startswith('#') or line.strip() == "":
+    continue
+
+  line = line.strip().split()
+  if not line[0].isdigit() or len(line) < 4:
+    continue
+
+  if start == -1:
+    start = float(line[2]) + warmup
+    end = start + warmup
+
+  fts = float(line[2])
+  
+  if fts < start:
+    continue
+
+  if fts > end:
+    break
+
+  latency = int(line[3])
+  status = int(line[4])
+  ttype = -1
+  try:
+    ttype = int(line[5])
+    extra = int(line[6])
+  except:
+    extra = 0
+
+  if status == 1 and ttype == 2:
+    xLatency.append(latency)
+
+  tLatency.append(latency) 
+  tExtra += extra
+
+  if status == 1:
+    sLatency.append(latency)
+    sExtra += extra
+  else:
+    fLatency.append(latency)
+    fExtra += extra
+
+if len(tLatency) == 0:
+  print "Zero completed transactions.."
+  sys.exit()
+
+tLatency.sort()
+sLatency.sort()
+fLatency.sort()
+
+print "Transactions(All/Success): ", len(tLatency), len(sLatency)
+print "Abort Rate: ", (float)(len(tLatency)-len(sLatency))/len(tLatency)
+print "Throughput (All/Success): ", len(tLatency)/(end-start), len(sLatency)/(end-start)
+print "Average Latency (all): ", sum(tLatency)/float(len(tLatency))
+print "Median  Latency (all): ", tLatency[len(tLatency)/2]
+print "Average Latency (success): ", sum(sLatency)/float(len(sLatency))
+print "Median  Latency (success): ", sLatency[len(sLatency)/2]
+print "Extra (all): ", tExtra
+print "Extra (success): ", sExtra
+if len(xLatency) > 0:
+  print "X Transaction Latency: ", sum(xLatency)/float(len(xLatency))
+if len(fLatency) > 0:
+  print "Average Latency (failure): ", sum(fLatency)/float(len(tLatency)-len(sLatency))
+  print "Extra (failure): ", fExtra

store/tools/run_test.sh

+#!/bin/bash
+
+trap '{
+  echo "\nKilling all clients.. Please wait..";
+  for host in ${clients[@]}
+  do
+    ssh $host "killall -9 $client";
+    ssh $host "killall -9 $client";
+  done
+
+  echo "\nKilling all replics.. Please wait..";
+  for host in ${servers[@]}
+  do
+    ssh $host "killall -9 server";
+  done
+}' INT
+
+# Paths to source code and logfiles.
+srcdir="/homes/sys/naveenks/Research/Tapir"
+logdir="/biggerraid/users/naveenks/tapir"
+
+# Machines on which replicas are running.
+replicas=("breakout" "pitfall" "qbert")
+
+# Machines on which clients are running.
+clients=("spyhunter")
+
+client="benchClient"    # Which client (benchClient, retwisClient, etc)
+store="tapirstore"      # Which store (strongstore, weakstore, tapirstore)
+mode="txn-l"            # Mode for storage system.
+
+nshard=1     # number of shards
+nclient=1    # number of clients to run (per machine)
+nkeys=100000 # number of keys to use
+rtime=10     # duration to run
+
+tlen=2       # transaction length
+wper=0       # writes percentage
+err=0        # error
+skew=0       # skew
+zalpha=-1    # zipf alpha (-1 to disable zipf and enable uniform)
+
+# Print out configuration being used.
+echo "Configuration:"
+echo "Shards: $nshard"
+echo "Clients per host: $nclient"
+echo "Threads per client: $nthread"
+echo "Keys: $nkeys"
+echo "Transaction Length: $tlen"
+echo "Write Percentage: $wper"
+echo "Error: $err"
+echo "Skew: $skew"
+echo "Zipf alpha: $zalpha"
+echo "Skew: $skew"
+echo "Client: $client"
+echo "Store: $store"
+echo "Mode: $mode"
+
+
+# Generate keys to be used in the experiment.
+echo "Generating random keys.."
+python key_generator.py $nkeys > keys
+
+
+# Start all replicas and timestamp servers
+echo "Starting TimeStampServer replicas.."
+$srcdir/store/tools/start_replica.sh tss $srcdir/store/tools/shard.tss.config \
+  "$srcdir/timeserver/timeserver" $logdir
+
+for ((i=0; i<$nshard; i++))
+do
+  echo "Starting shard$i replicas.."
+  $srcdir/store/tools/start_replica.sh shard$i $srcdir/store/tools/shard$i.config \
+    "$srcdir/store/$store/server -m $mode -f $srcdir/store/tools/keys -e $err -s $skew" $logdir
+done
+
+
+# Wait a bit for all replicas to start up
+sleep 2
+
+
+# Run the clients
+echo "Running the client(s)"
+count=0
+for host in ${clients[@]}
+do
+  ssh $host "$srcdir/store/tools/start_client.sh \"$srcdir/store/benchmark/$client \
+  -c $srcdir/store/tools/shard -N $nshard -f $srcdir/store/tools/keys \
+  -d $rtime -l $tlen -w $wper -k $nkeys -m $mode -e $err -s $skew -z $zalpha\" \
+  $count $nclient $logdir"
+
+  let count=$count+$nclient
+done
+
+
+# Wait for all clients to exit
+echo "Waiting for client(s) to exit"
+for host in ${clients[@]}
+do
+  ssh $host "$srcdir/store/tools/wait_client.sh $client"
+done
+
+
+# Kill all replicas
+echo "Cleaning up"
+$srcdir/store/tools/stop_replica.sh $srcdir/store/tools/shard.tss.config > /dev/null 2>&1
+for ((i=0; i<$nshard; i++))
+do
+  $srcdir/store/tools/stop_replica.sh $srcdir/store/tools/shard$i.config > /dev/null 2>&1
+done
+
+
+# Process logs
+echo "Processing logs"
+cat $logdir/client.*.log | sort -g -k 3 > $logdir/client.log
+rm -f $logdir/client.*.log
+
+python $srcdir/store/tools/process_logs.py $logdir/client.log $rtime

store/tools/shard.tss.config

+f 1
+replica breakout:51735
+replica pitfall:51735
+replica qbert:51735

store/tools/shard0.config

+f 1
+replica breakout:51729
+replica pitfall:51729
+replica qbert:51729

store/tools/shard1.config

+f 1
+replica breakout:51730
+replica pitfall:51730
+replica qbert:51730

store/tools/shard2.config

+f 1
+replica breakout:51731
+replica pitfall:51731
+replica qbert:51731

store/tools/shard3.config

+f 1
+replica breakout:51732
+replica pitfall:51732
+replica qbert:51732

store/tools/shard4.config

+f 1
+replica breakout:51733
+replica pitfall:51733
+replica qbert:51733

store/tools/start_client.sh

+#!/bin/bash
+
+if [ "$#" -ne 4 ]; then
+  echo "Usage: $0 command begin_id ncopies" >&2
+  exit 1
+fi
+
+cmd=$1
+begin=$2
+copies=$3
+logdir=$4
+
+let end=$begin+$copies
+
+for ((i=$begin; i<$end; i++))
+do
+  command="$cmd > $logdir/client.$i.log 2>&1 &"
+  #echo $command
+  eval $command
+done

store/tools/start_replica.sh

+#!/bin/bash
+
+shard=$1    # which shard is this
+config=$2   # path to config file
+cmd=$3      # command to run
+logdir=$4   # log directory
+
+if [ "$#" -ne 4 ]; then
+  echo "Usage: $0 shard configpath command logdir" >&2
+  exit 1
+fi
+
+n=$(head -n1 $config | awk '{print $2}')
+let n=2*$n+1
+
+for ((i=0; i<$n; i++))
+do
+  let line=$i+2 
+  server=$(cat $config | sed -n ${line}p | awk -F'[ :]' '{print $2}')
+  command="ssh $server \"$cmd -c $config -i $i > $logdir/$shard.replica$i.log 2>&1 &\""
+  #echo $command
+  eval $command
+done

store/tools/stop_replica.sh

+#!/bin/bash
+
+config=$1   # path to config file
+
+if [ "$#" -ne 1 ]; then
+  echo "Usage: $0 configpath" >&2
+  exit 1
+fi
+
+n=$(head -n1 $config | awk '{print $2}')
+let n=2*$n+1
+
+for ((i=0; i<$n; i++))
+do
+  let line=$i+2 
+  server=$(cat $config | sed -n ${line}p | awk -F'[ :]' '{print $2}')
+  command="ssh $server \"pkill -INT server\""
+  #echo $command
+  eval $command
+done

store/tools/wait_client.sh

+#/bin/bash
+
+procname=$1
+check=1
+while [ $check -gt 0 ]
+do
+    check=`pgrep -u $USER -x $procname | wc -l`
+    sleep 1
+done