diff --git a/dockerfiles/Dockerfile b/dockerfiles/Dockerfile index ca74bb8..c455fd9 100644 --- a/dockerfiles/Dockerfile +++ b/dockerfiles/Dockerfile @@ -1,18 +1,19 @@ FROM maven:3.6.0-jdk-11 WORKDIR /graph-lib # Download webgraph binary RUN curl -O http://webgraph.di.unimi.it/webgraph-big-3.5.0-bin.tar.gz RUN tar xvfz webgraph-big-3.5.0-bin.tar.gz RUN cp webgraph-big-3.5.0/webgraph-big-3.5.0.jar . # Download webgraph dependencies RUN curl -O http://webgraph.di.unimi.it/webgraph-big-deps.tar.gz RUN tar xvfz webgraph-big-deps.tar.gz # Monitoring RUN apt-get update RUN apt-get install -y time WORKDIR /graph +COPY configuration configuration/ COPY scripts scripts/ diff --git a/dockerfiles/configuration/logback.xml b/dockerfiles/configuration/logback.xml new file mode 100644 index 0000000..76a4c7a --- /dev/null +++ b/dockerfiles/configuration/logback.xml @@ -0,0 +1,11 @@ + + + + %d %r %p [%t] %logger{1} - %m%n + + + + + + + diff --git a/dockerfiles/scripts/compress_graph.sh b/dockerfiles/scripts/compress_graph.sh index 8ac2756..5b79cb3 100755 --- a/dockerfiles/scripts/compress_graph.sh +++ b/dockerfiles/scripts/compress_graph.sh @@ -1,86 +1,87 @@ #!/bin/bash usage() { echo "Usage: --input --output --lib " echo " options:" echo " -t, --tmp (default to /tmp/)" exit 1 } graph_path="" out_dir="" lib_path="" tmp_dir="/tmp/" while (( "$#" )); do case "$1" in -i|--input) shift; graph_path=$1;; -o|--output) shift; out_dir=$1;; -l|--lib) shift; lib_path=$1;; -t|--tmp) shift; tmp_dir=$1;; *) usage;; esac shift done if [[ -z $graph_path || -z $out_dir || -z $lib_path ]]; then usage fi dataset=$(basename $graph_path) compr_graph_path="$out_dir/$dataset" stdout_file="$out_dir/stdout" stderr_file="$out_dir/stderr" mkdir -p $out_dir mkdir -p $tmp_dir if [[ -f "$stdout_file" || -f "$stderr_file" ]]; then echo "Cannot overwrite compression stdout/stderr files" exit 1 fi java_cmd () { /usr/bin/time -v java \ -server -Xmx1024G -XX:PretenureSizeThreshold=512M \ -XX:MaxNewSize=4G -XX:+UseLargePages -XX:+UseNUMA \ -XX:+UseTransparentHugePages -XX:+UseTLAB -XX:+ResizeTLAB \ + -Dlogback.configurationFile=configuration/logback.xml \ -cp $lib_path/'*' $* } { # Build a function (MPH) that maps node names to node numbers in # lexicographic order (output: .mph) java_cmd it.unimi.dsi.sux4j.mph.GOVMinimalPerfectHashFunction \ --zipped $compr_graph_path.mph --temp-dir $tmp_dir \ $graph_path.nodes.csv.gz ; # Build the graph in BVGraph format (output: .{graph,offsets,properties}) java_cmd it.unimi.dsi.big.webgraph.ScatteredArcsASCIIGraph \ --function $compr_graph_path.mph --temp-dir $tmp_dir \ --zipped $compr_graph_path-bv < $graph_path.edges.csv.gz ; # Build the offset big-list file to load the graph faster (output: .obl) java_cmd it.unimi.dsi.big.webgraph.BVGraph \ --list $compr_graph_path-bv ; # Find a better permutation using a BFS traversal order (output: .order) java_cmd it.unimi.dsi.law.graph.BFSBig \ $compr_graph_path-bv $compr_graph_path.order ; # Permute the graph accordingly batch_size=1000000000 java_cmd it.unimi.dsi.big.webgraph.Transform mapOffline \ $compr_graph_path-bv $compr_graph_path \ $compr_graph_path.order $batch_size ; java_cmd it.unimi.dsi.big.webgraph.BVGraph \ --list $compr_graph_path ; # Compute graph statistics (output: .{indegree,outdegree,stats}) java_cmd it.unimi.dsi.big.webgraph.Stats $compr_graph_path ; # Create transposed graph (to allow backward traversal) java_cmd it.unimi.dsi.big.webgraph.Transform transposeOffline \ $compr_graph_path $compr_graph_path-transposed $batch_size ; java_cmd it.unimi.dsi.big.webgraph.BVGraph \ --list $compr_graph_path-transposed ; } >> $stdout_file 2>> $stderr_file echo "Graph compression done."