From 41ca9703fd05ee8d7c3736522fed2d4e5a660150 Mon Sep 17 00:00:00 2001 From: Erick Lavoie <erick.lavoie@epfl.ch> Date: Tue, 15 Mar 2022 14:28:15 +0100 Subject: [PATCH] Added option to disable dynamic allocation on the cluster for performance tests --- README.md | 2 ++ timeCluster.sh | 4 ++-- 2 files changed, 4 insertions(+), 2 deletions(-) diff --git a/README.md b/README.md index a9f1daa..6191a8b 100644 --- a/README.md +++ b/README.md @@ -123,6 +123,8 @@ On ````ml-25m````: You can vary the number of executors used locally by using ````local[X]```` with X being an integer representing the number of cores you want to use locally. +You can vary the number of executors on the cluster by disabling dynamic allocation with ````--conf "spark.dynamicAllocation.enabled=false"````, using ````--master yarn````, and setting the number of executors with ````--num-executors X````. + ### Personalized ```` diff --git a/timeCluster.sh b/timeCluster.sh index 597d3c3..c73d0ad 100755 --- a/timeCluster.sh +++ b/timeCluster.sh @@ -11,6 +11,6 @@ source ./config.sh echo "------------------- DISTRIBUTED ---------------------" >> $LOGS sbt assembly # 1 Executor -spark-submit --class distributed.DistributedBaseline --master $SPARKMASTER --num-executors 1 target/scala-2.11/m1_yourid-assembly-1.0.jar --train $ML25Mr2train --test $ML25Mr2test --separator , --json $RUN/distributed-25m-1.json --num_measurements 3 2>&1 >>$LOGS +spark-submit --class distributed.DistributedBaseline --master $SPARKMASTER --num-executors 1 --conf "spark.dynamicAllocation.enabled=false" target/scala-2.11/m1_yourid-assembly-1.0.jar --train $ML25Mr2train --test $ML25Mr2test --separator , --json $RUN/distributed-25m-1.json --num_measurements 3 2>&1 >>$LOGS # 4 Executors -spark-submit --class distributed.DistributedBaseline --master $SPARKMASTER --num-executors 4 target/scala-2.11/m1_yourid-assembly-1.0.jar --train $ML25Mr2train --test $ML25Mr2test --separator , --json $RUN/distributed-25m-4.json --num_measurements 3 2>&1 >>$LOGS +spark-submit --class distributed.DistributedBaseline --master $SPARKMASTER --num-executors 4 --conf "spark.dynamicAllocation.enabled=false" target/scala-2.11/m1_yourid-assembly-1.0.jar --train $ML25Mr2train --test $ML25Mr2test --separator , --json $RUN/distributed-25m-4.json --num_measurements 3 2>&1 >>$LOGS -- GitLab