TPC-H and TPC-DS performance benchmark for Spark with DeltaLake

Based on: Delta Lake: High-Performance ACID Table Storage over Cloud Object Stores

Table of Contents

1. Clone submodules
2. Prerequisites
   • Install Dependencies
   • Download spark
   • Download sbt launcher
3. Build
   • spark-sql-perf
   • tpcds-kit
   • tpch-dbgen
4. Performance Benchmarking
   • TPCH
     • Parquet
     • ORC
     • CSV
     • JSON
   • TPCDS
     • Parquet
     • ORC
     • CSV
     • JSON
5. Reports
   • TPCH Report
   • TPCDS Report

1. Clone submodules

git submodule init
git submodule update

2. Prerequisites

2.1 Install Dependencies

sudo apt-get install unzip zip gcc make flex bison byacc git build-essential -y
curl -s "https://get.sdkman.io" | bash
source "$HOME/.sdkman/bin/sdkman-init.sh"
sdk install java 8.0.302-open
sdk install sbt 0.13.18

2.2 Download spark

wget https://archive.apache.org/dist/spark/spark-3.2.0/spark-3.2.0-bin-hadoop3.2.tgz
mkdir spark3
tar -xzf spark-3.2.0-bin-hadoop3.2.tgz --strip 1 -C spark3

NOTE: Deploy spark in standalone cluster mode.

2.3 Download sbt launcher

wget -P /tmp https://github.com/sbt/sbt/releases/download/v0.13.18/sbt-0.13.18.tgz
tar -xf /tmp/sbt-0.13.18.tgz  -C /tmp

3. Build

3.1 spark-sql-perf

cd spark-sql-perf
cp /tmp/sbt/bin/sbt-launch.jar build/sbt-launch-0.13.18.jar
bin/run
sbt +package

3.2 tpcds-kit

cd ../tpcds-kit/tools
make CC=gcc-9 OS=LINUX

3.3 tpch-dbgen

cd ../../tpch-dbgen
git checkout 0469309147b42abac8857fa61b4cf69a6d3128a8 -- bm_utils.c
make

NOTE: This should be installed on all cluster nodes with the same location and build tpcds-kit, tpch-dbgen

4. Performance Benchmarking

NOTE: Change master-ip executor-memory , num-executors,executor-cores a/c to your machine specifications in .sh files.

4.1 TPCH

4.1.1 Parquet

cd ../tpch
#For generating ~100GB parquet data
./gendata_parquet.sh
# For runing all 22 TPC-H Queries
./runtpch_parquet.sh

4.1.2 ORC

cd ../tpch
#For generating ~100GB orc data
./gendata_orc.sh
# For runing all 22 TPC-H Queries
./runtpch_orc.sh

4.1.3 CSV

cd ../tpch
#For generating ~100GB csv data
./gendata_csv.sh
# For runing all 22 TPC-H Queries
./runtpch_csv.sh

4.1.4 JSON

cd ../tpch
#For generating ~100GB csv data
./gendata_json.sh
# For runing all 22 TPC-H Queries
./runtpch_son.sh

4.2 TPCDS

4.2.1 Parquet

cd ../tpcds
#For generating ~100GB parquet data
./gendata_parquet.sh
# For runing all 99 TPC-DS Queries
./runtpch_parquet.sh

4.2.2 ORC

cd ../tpcds
#For generating ~100GB orc data
./gendata_orc.sh
# For runing all 99 TPC-DS Queries
./runtpch_orc.sh

4.2.3 CSV

cd ../tpcds
#For generating ~100GB csv data
./gendata_csv.sh
# For runing all 99 TPC-DS Queries
./runtpch_csv.sh

4.2.4 JSON

cd ../tpcds
#For generating ~100GB csv data
./gendata_json.sh
# For runing all 99 TPC-DS Queries
./runtpch_json.sh

## 5. Reports

### 5.1 TPCH reports

```bash
cd tpch/tpch_<parquet,orc,csv>_reports
# result will be present in part*.csv file

5.2 TPCDS reports

cd tpcds/tpcds_<parquet,orc,csv,json>_reports
# result will be present in part*.csv file