configurable number of partitions used in the SparkCreateSimRels phase

This commit is contained in:
Claudio Atzori 2020-07-13 16:07:07 +02:00
parent c73168b18e
commit 8c67938ad0
4 changed files with 50 additions and 32 deletions

View File

@ -2,6 +2,7 @@
package eu.dnetlib.dhp.oa.dedup; package eu.dnetlib.dhp.oa.dedup;
import java.io.IOException; import java.io.IOException;
import java.util.Optional;
import org.apache.commons.io.IOUtils; import org.apache.commons.io.IOUtils;
import org.apache.spark.SparkConf; import org.apache.spark.SparkConf;
@ -34,7 +35,7 @@ public class SparkCreateSimRels extends AbstractSparkAction {
private static final Logger log = LoggerFactory.getLogger(SparkCreateSimRels.class); private static final Logger log = LoggerFactory.getLogger(SparkCreateSimRels.class);
public static final int NUM_PARTITIONS = 10000; public static final int NUM_PARTITIONS = 1000;
public SparkCreateSimRels(ArgumentApplicationParser parser, SparkSession spark) { public SparkCreateSimRels(ArgumentApplicationParser parser, SparkSession spark) {
super(parser, spark); super(parser, spark);
@ -63,7 +64,12 @@ public class SparkCreateSimRels extends AbstractSparkAction {
final String isLookUpUrl = parser.get("isLookUpUrl"); final String isLookUpUrl = parser.get("isLookUpUrl");
final String actionSetId = parser.get("actionSetId"); final String actionSetId = parser.get("actionSetId");
final String workingPath = parser.get("workingPath"); final String workingPath = parser.get("workingPath");
final int numPartitions = Optional
.ofNullable(parser.get("numPartitions"))
.map(Integer::valueOf)
.orElse(NUM_PARTITIONS);
log.info("numPartitions: '{}'", numPartitions);
log.info("graphBasePath: '{}'", graphBasePath); log.info("graphBasePath: '{}'", graphBasePath);
log.info("isLookUpUrl: '{}'", isLookUpUrl); log.info("isLookUpUrl: '{}'", isLookUpUrl);
log.info("actionSetId: '{}'", actionSetId); log.info("actionSetId: '{}'", actionSetId);
@ -83,7 +89,7 @@ public class SparkCreateSimRels extends AbstractSparkAction {
JavaPairRDD<String, MapDocument> mapDocuments = sc JavaPairRDD<String, MapDocument> mapDocuments = sc
.textFile(DedupUtility.createEntityPath(graphBasePath, subEntity)) .textFile(DedupUtility.createEntityPath(graphBasePath, subEntity))
.repartition(NUM_PARTITIONS) .repartition(numPartitions)
.mapToPair( .mapToPair(
(PairFunction<String, String, MapDocument>) s -> { (PairFunction<String, String, MapDocument>) s -> {
MapDocument d = MapDocumentUtil.asMapDocumentWithJPath(dedupConf, s); MapDocument d = MapDocumentUtil.asMapDocumentWithJPath(dedupConf, s);
@ -93,13 +99,13 @@ public class SparkCreateSimRels extends AbstractSparkAction {
// create blocks for deduplication // create blocks for deduplication
JavaPairRDD<String, Block> blocks = Deduper JavaPairRDD<String, Block> blocks = Deduper
.createSortedBlocks(mapDocuments, dedupConf) .createSortedBlocks(mapDocuments, dedupConf)
.repartition(NUM_PARTITIONS); .repartition(numPartitions);
// create relations by comparing only elements in the same group // create relations by comparing only elements in the same group
Deduper Deduper
.computeRelations(sc, blocks, dedupConf) .computeRelations(sc, blocks, dedupConf)
.map(t -> createSimRel(t._1(), t._2(), entity)) .map(t -> createSimRel(t._1(), t._2(), entity))
.repartition(NUM_PARTITIONS) .repartition(numPartitions)
.map(r -> OBJECT_MAPPER.writeValueAsString(r)) .map(r -> OBJECT_MAPPER.writeValueAsString(r))
.saveAsTextFile(outputPath); .saveAsTextFile(outputPath);
} }

View File

@ -22,5 +22,11 @@
"paramLongName": "workingPath", "paramLongName": "workingPath",
"paramDescription": "path of the working directory", "paramDescription": "path of the working directory",
"paramRequired": true "paramRequired": true
},
{
"paramName": "np",
"paramLongName": "numPartitions",
"paramDescription": "number of partitions for the similarity relations intermediate phases",
"paramRequired": false
} }
] ]

View File

@ -20,6 +20,10 @@
<name>dedupGraphPath</name> <name>dedupGraphPath</name>
<description>path for the output graph</description> <description>path for the output graph</description>
</property> </property>
<property>
<name>cutConnectedComponent</name>
<description>max number of elements in a connected component</description>
</property>
<property> <property>
<name>sparkDriverMemory</name> <name>sparkDriverMemory</name>
<description>memory for driver process</description> <description>memory for driver process</description>
@ -106,10 +110,11 @@
--conf spark.eventLog.dir=${nameNode}${spark2EventLogDir} --conf spark.eventLog.dir=${nameNode}${spark2EventLogDir}
--conf spark.sql.shuffle.partitions=3840 --conf spark.sql.shuffle.partitions=3840
</spark-opts> </spark-opts>
<arg>--i</arg><arg>${graphBasePath}</arg> <arg>--graphBasePath</arg><arg>${graphBasePath}</arg>
<arg>--la</arg><arg>${isLookUpUrl}</arg> <arg>--isLookUpUrl</arg><arg>${isLookUpUrl}</arg>
<arg>--asi</arg><arg>${actionSetId}</arg> <arg>--actionSetId</arg><arg>${actionSetId}</arg>
<arg>--w</arg><arg>${workingPath}</arg> <arg>--workingPath</arg><arg>${workingPath}</arg>
<arg>--numPartitions</arg><arg>8000</arg>
</spark> </spark>
<ok to="CreateMergeRel"/> <ok to="CreateMergeRel"/>
<error to="Kill"/> <error to="Kill"/>
@ -132,10 +137,11 @@
--conf spark.eventLog.dir=${nameNode}${spark2EventLogDir} --conf spark.eventLog.dir=${nameNode}${spark2EventLogDir}
--conf spark.sql.shuffle.partitions=3840 --conf spark.sql.shuffle.partitions=3840
</spark-opts> </spark-opts>
<arg>--i</arg><arg>${graphBasePath}</arg> <arg>--graphBasePath</arg><arg>${graphBasePath}</arg>
<arg>--w</arg><arg>${workingPath}</arg> <arg>--workingPath</arg><arg>${workingPath}</arg>
<arg>--la</arg><arg>${isLookUpUrl}</arg> <arg>--isLookUpUrl</arg><arg>${isLookUpUrl}</arg>
<arg>--asi</arg><arg>${actionSetId}</arg> <arg>--actionSetId</arg><arg>${actionSetId}</arg>
<arg>--cutConnectedComponent</arg><arg>${cutConnectedComponent}</arg>
</spark> </spark>
<ok to="CreateDedupRecord"/> <ok to="CreateDedupRecord"/>
<error to="Kill"/> <error to="Kill"/>
@ -158,10 +164,10 @@
--conf spark.eventLog.dir=${nameNode}${spark2EventLogDir} --conf spark.eventLog.dir=${nameNode}${spark2EventLogDir}
--conf spark.sql.shuffle.partitions=3840 --conf spark.sql.shuffle.partitions=3840
</spark-opts> </spark-opts>
<arg>--i</arg><arg>${graphBasePath}</arg> <arg>--graphBasePath</arg><arg>${graphBasePath}</arg>
<arg>--w</arg><arg>${workingPath}</arg> <arg>--workingPath</arg><arg>${workingPath}</arg>
<arg>--la</arg><arg>${isLookUpUrl}</arg> <arg>--isLookUpUrl</arg><arg>${isLookUpUrl}</arg>
<arg>--asi</arg><arg>${actionSetId}</arg> <arg>--actionSetId</arg><arg>${actionSetId}</arg>
</spark> </spark>
<ok to="UpdateEntity"/> <ok to="UpdateEntity"/>
<error to="Kill"/> <error to="Kill"/>
@ -184,9 +190,9 @@
--conf spark.eventLog.dir=${nameNode}${spark2EventLogDir} --conf spark.eventLog.dir=${nameNode}${spark2EventLogDir}
--conf spark.sql.shuffle.partitions=3840 --conf spark.sql.shuffle.partitions=3840
</spark-opts> </spark-opts>
<arg>--i</arg><arg>${graphBasePath}</arg> <arg>--graphBasePath</arg><arg>${graphBasePath}</arg>
<arg>--w</arg><arg>${workingPath}</arg> <arg>--workingPath</arg><arg>${workingPath}</arg>
<arg>--o</arg><arg>${dedupGraphPath}</arg> <arg>--dedupGraphPath</arg><arg>${dedupGraphPath}</arg>
</spark> </spark>
<ok to="copyRelations"/> <ok to="copyRelations"/>
<error to="Kill"/> <error to="Kill"/>

View File

@ -1,17 +1,17 @@
[ [
{ {
"paramName": "i", "paramName": "i",
"paramLongName": "graphBasePath", "paramLongName": "graphBasePath",
"paramDescription": "the base path of raw graph", "paramDescription": "the base path of raw graph",
"paramRequired": true "paramRequired": true
}, },
{ {
"paramName": "w", "paramName": "w",
"paramLongName": "workingPath", "paramLongName": "workingPath",
"paramDescription": "the working directory path", "paramDescription": "the working directory path",
"paramRequired": true "paramRequired": true
}, },
{ {
"paramName": "o", "paramName": "o",
"paramLongName": "dedupGraphPath", "paramLongName": "dedupGraphPath",
"paramDescription": "the path of the dedup graph", "paramDescription": "the path of the dedup graph",