Skip to content

Commit 94de560

Browse files
heary-caodongjoon-hyun
heary-cao
authored and
dongjoon-hyun
committed
[SPARK-25848][SQL][TEST] Refactor CSVBenchmarks to use main method
## What changes were proposed in this pull request? use spark-submit: `bin/spark-submit --class org.apache.spark.sql.execution.datasources.csv.CSVBenchmark --jars ./core/target/spark-core_2.11-3.0.0-SNAPSHOT-tests.jar,./sql/catalyst/target/spark-catalyst_2.11-3.0.0-SNAPSHOT-tests.jar ./sql/core/target/spark-sql_2.11-3.0.0-SNAPSHOT-tests.jar` Generate benchmark result: `SPARK_GENERATE_BENCHMARK_FILES=1 build/sbt "sql/test:runMain org.apache.spark.sql.execution.datasources.csv.CSVBenchmark"` ## How was this patch tested? manual tests Closes #22845 from heary-cao/CSVBenchmarks. Authored-by: caoxuewen <cao.xuewen@zte.com.cn> Signed-off-by: Dongjoon Hyun <dongjoon@apache.org>
1 parent a129f07 commit 94de560

File tree

2 files changed

+51
-46
lines changed

2 files changed

+51
-46
lines changed

‎sql/core/benchmarks/CSVBenchmark-results.txt‎

Lines changed: 27 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,27 @@
1+
================================================================================================
2+
Benchmark to measure CSV read/write performance
3+
================================================================================================
4+
5+
OpenJDK 64-Bit Server VM 1.8.0_191-b12 on Linux 3.10.0-862.3.2.el7.x86_64
6+
Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
7+
Parsing quoted values: Best/Avg Time(ms) Rate(M/s) Per Row(ns) Relative
8+
------------------------------------------------------------------------------------------------
9+
One quoted string 64733 / 64839 0.0 1294653.1 1.0X
10+
11+
OpenJDK 64-Bit Server VM 1.8.0_191-b12 on Linux 3.10.0-862.3.2.el7.x86_64
12+
Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
13+
Wide rows with 1000 columns: Best/Avg Time(ms) Rate(M/s) Per Row(ns) Relative
14+
------------------------------------------------------------------------------------------------
15+
Select 1000 columns 185609 / 189735 0.0 185608.6 1.0X
16+
Select 100 columns 50195 / 51808 0.0 50194.8 3.7X
17+
Select one column 39266 / 39293 0.0 39265.6 4.7X
18+
count() 10959 / 11000 0.1 10958.5 16.9X
19+
20+
OpenJDK 64-Bit Server VM 1.8.0_191-b12 on Linux 3.10.0-862.3.2.el7.x86_64
21+
Intel(R) Xeon(R) CPU E5-2670 v2 @ 2.50GHz
22+
Count a dataset with 10 columns: Best/Avg Time(ms) Rate(M/s) Per Row(ns) Relative
23+
------------------------------------------------------------------------------------------------
24+
Select 10 columns + count() 24637 / 24768 0.4 2463.7 1.0X
25+
Select 1 column + count() 20026 / 20076 0.5 2002.6 1.2X
26+
count() 3754 / 3877 2.7 375.4 6.6X
27+

sql/core/src/test/scala/org/apache/spark/sql/execution/datasources/csv/CSVBenchmarks.scala renamed to sql/core/src/test/scala/org/apache/spark/sql/execution/datasources/csv/CSVBenchmark.scala

Lines changed: 24 additions & 46 deletions
Original file line numberDiff line numberDiff line change
@@ -16,30 +16,31 @@
1616
*/
1717
package org.apache.spark.sql.execution.datasources.csv
1818

19-
import org.apache.spark.SparkConf
2019
import org.apache.spark.benchmark.Benchmark
21-
import org.apache.spark.sql.{Column, Row, SparkSession}
22-
import org.apache.spark.sql.catalyst.plans.SQLHelper
20+
import org.apache.spark.sql.{Column, Row}
21+
import org.apache.spark.sql.execution.benchmark.SqlBasedBenchmark
2322
import org.apache.spark.sql.functions.lit
2423
import org.apache.spark.sql.types._
2524

2625
/**
2726
* Benchmark to measure CSV read/write performance.
28-
* To run this:
29-
* spark-submit --class <this class> --jars <spark sql test jar>
27+
* To run this benchmark:
28+
* {{{
29+
* 1. without sbt:
30+
* bin/spark-submit --class <this class> --jars <spark core test jar>,
31+
* <spark catalyst test jar> <spark sql test jar>
32+
* 2. build/sbt "sql/test:runMain <this class>"
33+
* 3. generate result:
34+
* SPARK_GENERATE_BENCHMARK_FILES=1 build/sbt "sql/test:runMain <this class>"
35+
* Results will be written to "benchmarks/CSVBenchmark-results.txt".
36+
* }}}
3037
*/
31-
object CSVBenchmarks extends SQLHelper {
32-
val conf = new SparkConf()
33-
34-
val spark = SparkSession.builder
35-
.master("local[1]")
36-
.appName("benchmark-csv-datasource")
37-
.config(conf)
38-
.getOrCreate()
38+
39+
object CSVBenchmark extends SqlBasedBenchmark {
3940
import spark.implicits._
4041

4142
def quotedValuesBenchmark(rowsNum: Int, numIters: Int): Unit = {
42-
val benchmark = new Benchmark(s"Parsing quoted values", rowsNum)
43+
val benchmark = new Benchmark(s"Parsing quoted values", rowsNum, output = output)
4344

4445
withTempPath { path =>
4546
val str = (0 until 10000).map(i => s""""$i"""").mkString(",")
@@ -56,20 +57,13 @@ object CSVBenchmarks extends SQLHelper {
5657
ds.filter((_: Row) => true).count()
5758
}
5859

59-
/*
60-
Intel(R) Core(TM) i7-7920HQ CPU @ 3.10GHz
61-
62-
Parsing quoted values: Best/Avg Time(ms) Rate(M/s) Per Row(ns) Relative
63-
--------------------------------------------------------------------------------------------
64-
One quoted string 30273 / 30549 0.0 605451.2 1.0X
65-
*/
6660
benchmark.run()
6761
}
6862
}
6963

7064
def multiColumnsBenchmark(rowsNum: Int): Unit = {
7165
val colsNum = 1000
72-
val benchmark = new Benchmark(s"Wide rows with $colsNum columns", rowsNum)
66+
val benchmark = new Benchmark(s"Wide rows with $colsNum columns", rowsNum, output = output)
7367

7468
withTempPath { path =>
7569
val fields = Seq.tabulate(colsNum)(i => StructField(s"col$i", IntegerType))
@@ -98,23 +92,14 @@ object CSVBenchmarks extends SQLHelper {
9892
ds.count()
9993
}
10094

101-
/*
102-
Intel(R) Core(TM) i7-7920HQ CPU @ 3.10GHz
103-
104-
Wide rows with 1000 columns: Best/Avg Time(ms) Rate(M/s) Per Row(ns) Relative
105-
--------------------------------------------------------------------------------------------
106-
Select 1000 columns 81091 / 81692 0.0 81090.7 1.0X
107-
Select 100 columns 30003 / 34448 0.0 30003.0 2.7X
108-
Select one column 24792 / 24855 0.0 24792.0 3.3X
109-
count() 24344 / 24642 0.0 24343.8 3.3X
110-
*/
11195
benchmark.run()
11296
}
11397
}
11498

11599
def countBenchmark(rowsNum: Int): Unit = {
116100
val colsNum = 10
117-
val benchmark = new Benchmark(s"Count a dataset with $colsNum columns", rowsNum)
101+
val benchmark =
102+
new Benchmark(s"Count a dataset with $colsNum columns", rowsNum, output = output)
118103

119104
withTempPath { path =>
120105
val fields = Seq.tabulate(colsNum)(i => StructField(s"col$i", IntegerType))
@@ -137,22 +122,15 @@ object CSVBenchmarks extends SQLHelper {
137122
ds.count()
138123
}
139124

140-
/*
141-
Intel(R) Core(TM) i7-7700HQ CPU @ 2.80GHz
142-
143-
Count a dataset with 10 columns: Best/Avg Time(ms) Rate(M/s) Per Row(ns) Relative
144-
---------------------------------------------------------------------------------------------
145-
Select 10 columns + count() 12598 / 12740 0.8 1259.8 1.0X
146-
Select 1 column + count() 7960 / 8175 1.3 796.0 1.6X
147-
count() 2332 / 2386 4.3 233.2 5.4X
148-
*/
149125
benchmark.run()
150126
}
151127
}
152128

153-
def main(args: Array[String]): Unit = {
154-
quotedValuesBenchmark(rowsNum = 50 * 1000, numIters = 3)
155-
multiColumnsBenchmark(rowsNum = 1000 * 1000)
156-
countBenchmark(10 * 1000 * 1000)
129+
override def runBenchmarkSuite(mainArgs: Array[String]): Unit = {
130+
runBenchmark("Benchmark to measure CSV read/write performance") {
131+
quotedValuesBenchmark(rowsNum = 50 * 1000, numIters = 3)
132+
multiColumnsBenchmark(rowsNum = 1000 * 1000)
133+
countBenchmark(10 * 1000 * 1000)
134+
}
157135
}
158136
}

0 commit comments

Comments
 (0)