TPCx-BB源码分析

Copy from: 一篇文章看懂TPCx-BB(大数据基准测试工具）源码

　　TPCx-BB是大数据基准测试工具,它通过模拟零售商的30个应用场景，执行30个查询来衡量基于Hadoop的大数据系统的包括硬件和软件的性能。其中一些场景还用到了机器学习算法（聚类、线性回归等）。为了更好地了解被测试的系统的性能，需要对TPCx-BB整个测试流程深入了解。本文详细分析了整个TPCx-BB测试工具的源码，希望能够对大家理解TPCx-BB有所帮助。

代码结构

主目录（$BENCH_MARK_HOME）下有：

  • bin
  • conf
  • data-generator
  • engines
  • tools

几个子目录。

bin下有几个 module ,是执行时需要用到的脚本：bigBench、cleanLogs、logEnvInformation、runBenchmark、zipLogs等

conf下有两个配置文件：bigBench.properties 和 userSettings.conf

bigBench.properties 主要设置 workload（执行的benchmarkPhases）和 power_test_0（POWER_TEST 阶段需要执行的SQL查询）

默认 workload ：

workload=CLEAN_ALL,ENGINE_VALIDATION_DATA_GENERATION,ENGINE_VALIDATION_LOAD_TEST,ENGINE_VALIDATION_POWER_TEST,ENGINE_VALIDATION_RESULT_VALIDATION,CLEAN_DATA,DATA_GENERATION,BENCHMARK_START,LOAD_TEST,POWER_TEST,THROUGHPUT_TEST_1,BENCHMARK_STOP,VALIDATE_POWER_TEST,VALIDATE_THROUGHPUT_TEST_1

默认 power_test_0 ：1-30

userSetting.conf 是一些基本设置，包括JAVA environment 、default settings for benchmark（database、engine、map_tasks、scale_factor ...）、HADOOP environment、
HDFS config and paths、Hadoop data generation options(DFS_REPLICATION、HADOOP_JVM_ENV...)

data-generator下是跟数据生成相关的脚本及配置文件。详细内容在下面介绍。

engines下是TPCx-BB支持的4种引擎：biginsights、hive、impala、spark_sql。默认引擎为hive。实际上，只有hive目录下不为空，其他三个目录下均为空，估计是现在还未完善。

tools下有两个jar包：HadoopClusterExec.jar 和 RunBigBench.jar 。其中 RunBigBench.jar 是执行TPCx-BB测试的一个非常重要的文件，大部分程序都在该jar包内。

数据生成

数据生成相关程序和配置都在 data-generator 目录下。该目录下有一个 pdgf.jar 包和 config、dicts、extlib 三个子目录。

pdgf.jar是数据生成的Java程序。config下有两个配置文件：bigbench-generation.xml 和 bigbench-schema.xml 。

bigbench-generation.xml 主要设置生成的原始数据（不是数据库表）包含哪几张表、每张表的表名、表的大小以及表输出的目录、表文件的后缀、分隔符、字符编码等。

<schema name="default">
        <tables>
        <!-- not refreshed tables -->    

            <!-- tables not used in benchmark, but some tables have references to them. not refreshed. Kept for legacy reasons  -->
            <table name="income_band"></table>
            <table name="reason"></table>
            <table name="ship_mode"></table>
            <table name="web_site"></table>
            <!-- /tables not used in benchmark  -->

            <!-- Static tables (fixed small size, generated only on node 1, skipped on others, not generated during refresh) -->
            <table name="date_dim" static="true"></table>
            <table name="time_dim" static="true"></table>
            <table name="customer_demographics" static="true"></table>
            <table name="household_demographics" static="true"></table>
            <!-- /static tables -->

            <!-- "normal" tables. split over all nodes. not generated during refresh -->
            <table name="store"></table>
            <table name="warehouse"></table>
            <table name="promotion"></table>
            <table name="web_page"></table>
            <!-- /"normal" tables.-->

        <!-- /not refreshed tables -->    

            <!--
            refreshed tables. Generated on all nodes.
            Refresh tables generate extra data during refresh (e.g. add new data to the existing tables)
            In "normal"-Phase  generate table rows:  [0,REFRESH_PERCENTAGE*Table.Size];
            In "refresh"-Phase generate table rows:  [REFRESH_PERCENTAGE*Table.Size+1, Table.Size]
            .Has effect only if  ${REFRESH_SYSTEM_ENABLED}==1.
            -->
            <table name="customer">
                <scheduler name="DefaultScheduler">
                    <partitioner
                        name="pdgf.core.dataGenerator.scheduler.TemplatePartitioner">
                        <prePartition><![CDATA[
                    if(${REFRESH_SYSTEM_ENABLED}>0){
                        int tableID = table.getTableID();
                        int timeID = 0;
                        long lastTableRow=table.getSize()-1;
                        long rowStart;
                        long rowStop;
                        boolean exclude=false;
                        long refreshRows=table.getSize()*(1.0-${REFRESH_PERCENTAGE});
                        if(${REFRESH_PHASE}>0){
                            //Refresh part
                            rowStart = lastTableRow - refreshRows +1;
                            rowStop  = lastTableRow;
                            if(refreshRows<=0){
                                exclude=true;
                            }

                        }else{
                            //"normal" part
                            rowStart = 0;
                            rowStop = lastTableRow - refreshRows;
                        }
                        return new pdgf.core.dataGenerator.scheduler.Partition(tableID, timeID,rowStart,rowStop,exclude);
                    }else{
                        //DEFAULT
                        return getParentPartitioner().getDefaultPrePartition(project, table);
                    }

                    ]]></prePartition>
                    </partitioner>
                </scheduler>
            </table>
<output name="SplitFileOutputWrapper">
  <!-- DEFAULT output for all Tables, if no table specific output is specified-->
    <output name="CSVRowOutput">
      <fileTemplate><![CDATA[outputDir + table.getName() +(nodeCount!=1?"_"+pdgf.util.StaticHelper.zeroPaddedNumber(nodeNumber,nodeCount):"")+ fileEnding]]></fileTemplate>
      <outputDir>output/</outputDir>
      <fileEnding>.dat</fileEnding>
      <delimiter>|</delimiter>
      <charset>UTF-8</charset>
      <sortByRowID>true</sortByRowID>
    </output>

    <output name="StatisticsOutput" active="1">
      <size>${item_size}</size><!-- a counter per item .. initialize later-->

      <fileTemplate><![CDATA[outputDir + table.getName()+"_audit" +(nodeCount!=1?"_"+pdgf.util.StaticHelper.zeroPaddedNumber(nodeNumber,nodeCount):"")+ fileEnding]]></fileTemplate>
      <outputDir>output/</outputDir>
      <fileEnding>.csv</fileEnding>
      <delimiter>,</delimiter>
      <header><!--"" + pdgf.util.Constants.DEFAULT_LINESEPARATOR-->
      </header>
      <footer></footer>

bigbench-schema.xml 设置了很多参数，有跟表的规模有关的，比如每张表的大小（记录的条数）;绝大多数是跟表的字段有关的，比如时间的起始、结束、性别比例、结婚比例、指标的上下界等。还具体定义了每个字段是怎么生成的，以及限制条件。示例如下：

生成的数据大小由 SCALE_FACTOR（-f） 决定。如 -f 1，则生成的数据总大小约为1G；-f 100，则生成的数据总大小约为100G。那么SCALE_FACTOR（-f） 是怎么精确控制生成的数据的大小呢？

原因是 SCALE_FACTOR（-f） 决定了每张表的记录数。如下，customer 表的记录数为 100000.0d * ${SF_sqrt}，即如果 -f 1 则 customer 表的记录数为 100000*sqrt(1)= 10万条 ;如果 -f 100 则 customer 表的记录数为 100000*sqrt(100)= 100万条

<property name="${customer_size}" type="long">100000.0d * ${SF_sqrt}</property>
<property name="${DIMENSION_TABLES_START_DAY}" type="datetime">2000-01-03 00:00:00</property>
<property name="${DIMENSION_TABLES_END_DAY}" type="datetime">2004-01-05 00:00:00</property>
<property name="${gender_likelihood}" type="double">0.5</property>
<property name="${married_likelihood}" type="double">0.3</property>
<property name="${WP_LINK_MIN}" type="double">2</property>
<property name="${WP_LINK_MAX}" type="double">25</property>
  <field name="d_date" size="13" type="CHAR" primary="false">
   <gen_DateTime>
     <disableRng>true</disableRng>
     <useFixedStepSize>true</useFixedStepSize>
     <startDate>${date_dim_begin_date}</startDate>
     <endDate>${date_dim_end_date}</endDate>
     <outputFormat>yyyy-MM-dd</outputFormat>
    </gen_DateTime>
  </field>
  <field name="t_time_id" size="16" type="CHAR" primary="false">
   <gen_ConvertNumberToString>
    <gen_Id/>
    <size>16.0</size>
    <characters>ABCDEFGHIJKLMNOPQRSTUVWXYZ</characters>
   </gen_ConvertNumberToString>
  </field>
<field name="cd_dep_employed_count" size="10" type="INTEGER" primary="false">
   <gen_Null probability="${NULL_CHANCE}">
   <gen_WeightedListItem filename="dicts/bigbench/ds-genProbabilities.txt" list="dependent_count" valueColumn="0" weightColumn="0" />
   </gen_Null>
  </field>

dicts下有city.dict、country.dict、male.dict、female.dict、state.dict、mail_provider.dict等字典文件，表里每一条记录的各个字段应该是从这些字典里生成的。

extlib下是引用的外部程序jar包。有 lucene-core-4.9.0.jar、commons-net-3.3.jar、xml-apis.jar和log4j-1.2.15.jar等

总结：

pdgf.jar根据bigbench-generation.xml 和 bigbench-schema.xml两个文件里的配置（表名、字段名、表的记录条数、每个字段生成的规则），从 dicts 目录下对应的 .dict
文件获取表中每一条记录、每个字段的值，生成原始数据。

customer 表里的某条记录如下：

0 AAAAAAAAAAAAAAAA 1824793 3203 2555 28776 14690 Ms. Marisa Harrington N 17 4 1988 UNITED ARAB EMIRATES RRCyuY3XfE3a [email protected]   gdMmGdU9

如果执行 TPCx-BB 测试时指定 -f 1（SCALE_FACTOR = 1） 则最终生成的原始数据总大小约为 1G(977M+8.6M)

[[email protected] ~]# hdfs dfs -du -h /user/root/benchmarks/bigbench/data
12.7 M   38.0 M   /user/root/benchmarks/bigbench/data/customer
5.1 M    15.4 M   /user/root/benchmarks/bigbench/data/customer_address
74.2 M   222.5 M  /user/root/benchmarks/bigbench/data/customer_demographics
14.7 M   44.0 M   /user/root/benchmarks/bigbench/data/date_dim
151.5 K  454.4 K  /user/root/benchmarks/bigbench/data/household_demographics
327      981      /user/root/benchmarks/bigbench/data/income_band
405.3 M  1.2 G    /user/root/benchmarks/bigbench/data/inventory
6.5 M    19.5 M   /user/root/benchmarks/bigbench/data/item
4.0 M    12.0 M   /user/root/benchmarks/bigbench/data/item_marketprices
53.7 M   161.2 M  /user/root/benchmarks/bigbench/data/product_reviews
45.3 K   135.9 K  /user/root/benchmarks/bigbench/data/promotion
3.0 K    9.1 K    /user/root/benchmarks/bigbench/data/reason
1.2 K    3.6 K    /user/root/benchmarks/bigbench/data/ship_mode
3.3 K    9.9 K    /user/root/benchmarks/bigbench/data/store
4.1 M    12.4 M   /user/root/benchmarks/bigbench/data/store_returns
88.5 M   265.4 M  /user/root/benchmarks/bigbench/data/store_sales
4.9 M    14.6 M   /user/root/benchmarks/bigbench/data/time_dim
584      1.7 K    /user/root/benchmarks/bigbench/data/warehouse
170.4 M  511.3 M  /user/root/benchmarks/bigbench/data/web_clickstreams
7.9 K    23.6 K   /user/root/benchmarks/bigbench/data/web_page
5.1 M    15.4 M   /user/root/benchmarks/bigbench/data/web_returns
127.6 M  382.8 M  /user/root/benchmarks/bigbench/data/web_sales
8.6 K    25.9 K   /user/root/benchmarks/bigbench/data/web_site

执行流程

要执行TPCx-BB测试，首先需要切换到TPCx-BB源程序的目录下，然后进入bin目录，执行以下语句：

./bigBench runBenchmark -f 1 -m 8 -s 2 -j 5

其中，-f、-m、-s、-j都是参数，用户可根据集群的性能以及自己的需求来设置。如果不指定，则使用默认值，默认值在 conf 目录下的 userSetting.conf 文件指定，如下：

export BIG_BENCH_DEFAULT_DATABASE="bigbench"
export BIG_BENCH_DEFAULT_ENGINE="hive"
export BIG_BENCH_DEFAULT_MAP_TASKS="80"
export BIG_BENCH_DEFAULT_SCALE_FACTOR="1000"
export BIG_BENCH_DEFAULT_NUMBER_OF_PARALLEL_STREAMS="2"
export BIG_BENCH_DEFAULT_BENCHMARK_PHASE="run_query"

默认 MAP_TASKS 为 80（-m 80）、SCALE_FACTOR 为 1000（-f 1000）、NUMBER_OF_PARALLEL_STREAMS 为 2（-s 2）。

所有可选参数及其意义如下：

General options:
-d  使用的数据库 (默认: $BIG_BENCH_DEFAULT_DATABASE -> bigbench)
-e  使用的引擎 (默认: $BIG_BENCH_DEFAULT_ENGINE -> hive)
-f  数据集的规模因子（scale factor） (默认: $BIG_BENCH_DEFAULT_SCALE_FACTOR -> 1000)
-h  显示帮助
-m  数据生成的`map tasks`数 (default: $BIG_BENCH_DEFAULT_MAP_TASKS)"
-s  并行的`stream`数 (默认: $BIG_BENCH_DEFAULT_NUMBER_OF_PARALLEL_STREAMS -> 2)

Driver specific options:
-a  伪装模式执行
-b  执行期间将调用的bash脚本在标准输出中打印出来
-i  指定需要执行的阶段 (详情见$BIG_BENCH_CONF_DIR/bigBench.properties)
-j  指定需要执行的查询 (默认：1-30共30个查询均执行)"
-U  解锁专家模式

若指定了-U,即解锁了专家模式，则：

echo "EXPERT MODE ACTIVE"
echo "WARNING - INTERNAL USE ONLY:"
echo "Only set manually if you know what you are doing!"
echo "Ignoring them is probably the best solution"
echo "Running individual modules:"
echo "Usage: `basename $0` module [options]"

-D  指定需要debug的查询部分. 大部分查询都只有一个单独的部分
-p  需要执行的benchmark phase (默认: $BIG_BENCH_DEFAULT_BENCHMARK_PHASE -> run_query)"
-q  指定需要执行哪个查询（只能指定一个）
-t  指定执行该查询时用第哪个stream
-v  metastore population的sql脚本 (默认: ${USER_POPULATE_FILE:-"$BIG_BENCH_POPULATION_DIR/hiveCreateLoad.sql"})"
-w  metastore refresh的sql脚本 (默认: ${USER_REFRESH_FILE:-"$BIG_BENCH_REFRESH_DIR/hiveRefreshCreateLoad.sql"})"
-y  含额外的用户自定义查询参数的文件 (global: $BIG_BENCH_ENGINE_CONF_DIR/queryParameters.sql)"
-z  含额外的用户自定义引擎设置的文件 (global: $BIG_BENCH_ENGINE_CONF_DIR/engineSettings.sql)"

List of available modules:
    $BIG_BENCH_ENGINE_BIN_DIR

回到刚刚执行TPCx-BB测试的语句：

./bigBench runBenchmark -f 1 -m 8 -s 2 -j 5

bigBench

bigBench是主脚本，runBenchmark是module。

bigBench 里设置了很多环境变量（包括路径、引擎、STREAM数等等），因为后面调用 runBigBench.jar 的时候需要在Java程序里读取这些环境变量。

bigBench 前面都是在做一些基本工作，如设置环境变量、解析用户输入参数、赋予文件权限、设置路径等等。到最后一步调用 runBenchmark 的 runModule() 方法：

  1. 设置基本路径

     export BIG_BENCH_VERSION="1.0"
 export BIG_BENCH_BIN_DIR="$BIG_BENCH_HOME/bin"
 export BIG_BENCH_CONF_DIR="$BIG_BENCH_HOME/conf"
 export BIG_BENCH_DATA_GENERATOR_DIR="$BIG_BENCH_HOME/data-generator"
 export BIG_BENCH_TOOLS_DIR="$BIG_BENCH_HOME/tools"
 export BIG_BENCH_LOGS_DIR="$BIG_BENCH_HOME/logs"
  2. 指定 core-site.xml 和 hdfs-site.xml 的路径

    数据生成时要用到Hadoop集群，生成在hdfs上

     export BIG_BENCH_DATAGEN_CORE_SITE="$BIG_BENCH_HADOOP_CONF/core-site.xml"
export BIG_BENCH_DATAGEN_HDFS_SITE="$BIG_BENCH_HADOOP_CONF/hdfs-site.xml"
  3. 赋予整个包下所有可执行文件权限（.sh/.jar/.py） 
     find "$BIG_BENCH_HOME" -name ‘*.sh‘ -exec chmod 755 {} +
find "$BIG_BENCH_HOME" -name ‘*.jar‘ -exec chmod 755 {} +
find "$BIG_BENCH_HOME" -name ‘*.py‘ -exec chmod 755 {} +
  4. 设置 userSetting.conf 的路径并 source 
     USER_SETTINGS="$BIG_BENCH_CONF_DIR/userSettings.conf"
 if [ ! -f "$USER_SETTINGS" ]
 then
   echo "User settings file $USER_SETTINGS not found"
   exit 1
 else
   source "$USER_SETTINGS"
 fi
  5. 解析输入参数和选项并根据选项的内容作设置

    第一个参数必须是module_name

    如果没有输入参数或者第一个参数以"-"开头，说明用户没有输入需要运行的module。

     if [[ $# -eq 0 || "`echo "$1" | cut -c1`" = "-" ]]
 then
   export MODULE_NAME=""
   SHOW_HELP="1"
 else
   export MODULE_NAME="$1"
   shift
 fi
 export LIST_OF_USER_OPTIONS="[email protected]"

    解析用户输入的参数

    根据用户输入的参数来设置环境变量

     while getopts ":d:D:e:f:hm:p:q:s:t:Uv:w:y:z:abi:j:" OPT; do
case "$OPT" in
 # script options
 d)
   #echo "-d was triggered, Parameter: $OPTARG" >&2
   USER_DATABASE="$OPTARG"
 ;;
 D)
   #echo "-D was triggered, Parameter: $OPTARG" >&2
   DEBUG_QUERY_PART="$OPTARG"
 ;;
 e)
   #echo "-e was triggered, Parameter: $OPTARG" >&2
   USER_ENGINE="$OPTARG"
 ;;
 f)
   #echo "-f was triggered, Parameter: $OPTARG" >&2
   USER_SCALE_FACTOR="$OPTARG"
 ;;
 h)
   #echo "-h was triggered, Parameter: $OPTARG" >&2
   SHOW_HELP="1"
 ;;
 m)
   #echo "-m was triggered, Parameter: $OPTARG" >&2
   USER_MAP_TASKS="$OPTARG"
 ;;
 p)
   #echo "-p was triggered, Parameter: $OPTARG" >&2
   USER_BENCHMARK_PHASE="$OPTARG"
 ;;
 q)
   #echo "-q was triggered, Parameter: $OPTARG" >&2
   QUERY_NUMBER="$OPTARG"
 ;;
 s)
   #echo "-t was triggered, Parameter: $OPTARG" >&2
   USER_NUMBER_OF_PARALLEL_STREAMS="$OPTARG"
 ;;
 t)
   #echo "-s was triggered, Parameter: $OPTARG" >&2
   USER_STREAM_NUMBER="$OPTARG"
 ;;
 U)
   #echo "-U was triggered, Parameter: $OPTARG" >&2
   USER_EXPERT_MODE="1"
 ;;
 v)
   #echo "-v was triggered, Parameter: $OPTARG" >&2
   USER_POPULATE_FILE="$OPTARG"
 ;;
 w)
   #echo "-w was triggered, Parameter: $OPTARG" >&2
   USER_REFRESH_FILE="$OPTARG"
 ;;
 y)
   #echo "-y was triggered, Parameter: $OPTARG" >&2
   USER_QUERY_PARAMS_FILE="$OPTARG"
 ;;
 z)
   #echo "-z was triggered, Parameter: $OPTARG" >&2
   USER_ENGINE_SETTINGS_FILE="$OPTARG"
 ;;
 # driver options
 a)
   #echo "-a was triggered, Parameter: $OPTARG" >&2
   export USER_PRETEND_MODE="1"
 ;;
 b)
   #echo "-b was triggered, Parameter: $OPTARG" >&2
   export USER_PRINT_STD_OUT="1"
 ;;
 i)
   #echo "-i was triggered, Parameter: $OPTARG" >&2
   export USER_DRIVER_WORKLOAD="$OPTARG"
 ;;
 j)
   #echo "-j was triggered, Parameter: $OPTARG" >&2
   export USER_DRIVER_QUERIES_TO_RUN="$OPTARG"
 ;;
 \?)
   echo "Invalid option: -$OPTARG" >&2
   exit 1
 ;;
 :)
   echo "Option -$OPTARG requires an argument." >&2
   exit 1
;;
esac
done

    设置全局变量。如果用户指定了某个参数的值，则采用该值，否则使用默认值。

     export BIG_BENCH_EXPERT_MODE="${USER_EXPERT_MODE:-"0"}"
export SHOW_HELP="${SHOW_HELP:-"0"}"
export BIG_BENCH_DATABASE="${USER_DATABASE:-"$BIG_BENCH_DEFAULT_DATABASE"}"
export BIG_BENCH_ENGINE="${USER_ENGINE:-"$BIG_BENCH_DEFAULT_ENGINE"}"
export BIG_BENCH_MAP_TASKS="${USER_MAP_TASKS:-"$BIG_BENCH_DEFAULT_MAP_TASKS"}"
export BIG_BENCH_SCALE_FACTOR="${USER_SCALE_FACTOR:-"$BIG_BENCH_DEFAULT_SCALE_FACTOR"}"
export BIG_BENCH_NUMBER_OF_PARALLEL_STREAMS="${USER_NUMBER_OF_PARALLEL_STREAMS:-"$BIG_BENCH_DEFAULT_NUMBER_OF_PARALLEL_STREAMS"}"
export BIG_BENCH_BENCHMARK_PHASE="${USER_BENCHMARK_PHASE:-"$BIG_BENCH_DEFAULT_BENCHMARK_PHASE"}"
export BIG_BENCH_STREAM_NUMBER="${USER_STREAM_NUMBER:-"0"}"
export BIG_BENCH_ENGINE_DIR="$BIG_BENCH_HOME/engines/$BIG_BENCH_ENGINE"
export BIG_BENCH_ENGINE_CONF_DIR="$BIG_BENCH_ENGINE_DIR/conf"
  6. 检测 -s -m -f -j的选项是否为数字 
     if [ -n "`echo "$BIG_BENCH_MAP_TASKS" | sed -e ‘s/[0-9]*//g‘`" ]
 then
   echo "$BIG_BENCH_MAP_TASKS is not a number"
 fi
 if [ -n "`echo "$BIG_BENCH_SCALE_FACTOR" | sed -e ‘s/[0-9]*//g‘`" ]
 then
   echo "$BIG_BENCH_SCALE_FACTOR is not a number"
 fi
 if [ -n "`echo "$BIG_BENCH_NUMBER_OF_PARALLEL_STREAMS" | sed -e ‘s/[0-9]*//g‘`" ]
 then
   echo "$BIG_BENCH_NUMBER_OF_PARALLEL_STREAMS is not a number"
 fi
 if [ -n "`echo "$BIG_BENCH_STREAM_NUMBER" | sed -e ‘s/[0-9]*//g‘`" ]
 then
   echo "$BIG_BENCH_STREAM_NUMBER is not a number"
 fi
  7. 检查引擎是否存在 
     if [ ! -d "$BIG_BENCH_ENGINE_DIR" ]
 then
   echo "Engine directory $BIG_BENCH_ENGINE_DIR not found. Aborting script..."
   exit 1
 fi
 if [ ! -d "$BIG_BENCH_ENGINE_CONF_DIR" ]
 then
   echo "Engine configuration directory $BIG_BENCH_ENGINE_CONF_DIR not found. Aborting script..."
   exit 1
 fi
  8. 设置 engineSetting.conf 路径并 source 
     ENGINE_SETTINGS="$BIG_BENCH_ENGINE_CONF_DIR/engineSettings.conf"
 if [ ! -f "$ENGINE_SETTINGS" ]
 then
   echo "Engine settings file $ENGINE_SETTINGS not found"
   exit 1
 else
   source "$ENGINE_SETTINGS"
 fi
  9. 检查module是否存在

    当输入某个module时，系统会先到$BIG_BENCH_ENGINE_BIN_DIR/目录下去找该module是否存在，如果存在，就source "$MODULE"；如果该目录下不存在指定的module，再到export MODULE="$BIG_BENCH_BIN_DIR/"目录下找该module，如果存在，就source "$MODULE"；否则，输出Module $MODULE not found, aborting script.

    export MODULE="$BIG_BENCH_ENGINE_BIN_DIR/$MODULE_NAME"
if [ -f "$MODULE" ]
then
  source "$MODULE"
else
  export MODULE="$BIG_BENCH_BIN_DIR/$MODULE_NAME"
  if [ -f "$MODULE" ]
  then
    source "$MODULE"
  else
    echo "Module $MODULE not found, aborting script."
    exit 1
  fi
fi
  10. 检查module里的runModule（）、helpModule ( )、runEngineCmd()方法是否有定义 
    MODULE_RUN_METHOD="runModule"
if ! declare -F "$MODULE_RUN_METHOD" > /dev/null 2>&1
then
  echo "$MODULE_RUN_METHOD was not implemented, aborting script"
  exit 1
fi
  11. 运行module

    如果module是runBenchmark，执行
    runCmdWithErrorCheck "$MODULE_RUN_METHOD"
    也就是
    runCmdWithErrorCheck runModule（）

由上可以看出，bigBench脚本主要执行一些如设置环境变量、赋予权限、检查并解析输入参数等基础工作，最终调用runBenchmark的runModule()方法继续往下执行。

runBenchmark

接下来看看runBenchmark脚本。

runBenchmark里有两个函数：helpModule ()和runModule ()。

helpModule ()就是显示帮助。

runModule ()是运行runBenchmark模块时真正调用的函数。该函数主要做四件事：

  1. 清除之前生成的日志
  2. 调用RunBigBench.jar来执行
  3. logEnvInformation
  4. 将日志文件夹打包成zip

源码如下：

runModule () {
  #check input parameters
  if [ "$BIG_BENCH_NUMBER_OF_PARALLEL_STREAMS" -le 0 ]
  then
    echo "The number of parallel streams -s must be greater than 0"
    return 1
  fi

  "${BIG_BENCH_BIN_DIR}/bigBench" cleanLogs -U $LIST_OF_USER_OPTIONS
  "$BIG_BENCH_JAVA" -jar "${BIG_BENCH_TOOLS_DIR}/RunBigBench.jar"
  "${BIG_BENCH_BIN_DIR}/bigBench" logEnvInformation -U $LIST_OF_USER_OPTIONS
  "${BIG_BENCH_BIN_DIR}/bigBench" zipLogs -U $LIST_OF_USER_OPTIONS
  return $?
}

相当于运行runBenchmark模块时又调用了cleanLogs、logEnvInformation、zipLogs三个模块以及RunBigBench.jar。其中RunBigBench.jar是TCPx-BB测试执行的核心代码，用Java语言编写。接下来分析RunBigBench.jar源码。

runModule()

runModule()函数用来执行某个module。我们已知，执行某个module需要切换到主目录下的bin目录，然后执行：

./bigBench module_name arguments

在runModule()函数里，cmdLine用来生成如上命令。

ArrayList cmdLine = new ArrayList();
cmdLine.add("bash");
cmdLine.add(this.runScript);
cmdLine.add(benchmarkPhase.getRunModule());
cmdLine.addAll(arguments);

其中，this.runScript为：

this.runScript = (String)env.get("BIG_BENCH_BIN_DIR") + "/bigBench";

benchmarkPhase.getRunModule()用来获得需要执行的module。

arguments为用户输入的参数。

至此，cmdLine为：

bash $BIG_BENCH_BIN_DIR/bigBench module_name arguments

那么，怎么让系统执行该bash命令呢？答案是调用runCmd()方法。

boolean successful = this.runCmd(this.homeDir, benchmarkPhase.isPrintStdOut(), (String[])cmdLine.toArray(new String[0]));

接下来介绍rumCmd()方法

runCmd()

runCmd()方法通过ProcessBuilder来创建一个操作系统进程，并用该进程执行以上的bash命令。

ProcessBuilder还可以设置工作目录和环境。

ProcessBuilder pb = new ProcessBuilder(command);
pb.directory(new File(workingDirectory));
Process p = null;
---
p = pb.start();

getQueryList()

getQueryList()用来获得需要执行的查询列表。从$BIG_BENCH_LOGS_DIR/bigBench.properties文件中读取。与$BIG_BENCH_HOME/conf/bigBench.properties内容一致。

bigBench.properties里power_test_0=1-30规定了powter_test_0阶段需要执行的查询及其顺序。

可以用区间如 5-12 或者单个数字如 21 表示，中间用 , 隔开。

power_test_0=28-25,2-5,10,22,30

表示powter_test_0阶段需要执行的查询及其顺序为：28,27,26,25,2,3,4,5,10,22,30

如果想让30个查询按顺序执行，则：

power_test_0=1-30

获得查询列表的源码如下：

    private List<Integer> getQueryList(BigBench.BenchmarkPhase benchmarkPhase, int streamNumber) {
        String SHUFFLED_NAME_PATTERN = "shuffledQueryList";
        BigBench.BenchmarkPhase queryOrderBasicPhase = BigBench.BenchmarkPhase.POWER_TEST;
        String propertyKey = benchmarkPhase.getQueryListProperty(streamNumber);
        boolean queryOrderCached = benchmarkPhase.isQueryOrderCached();
        if(queryOrderCached && this.queryListCache.containsKey(propertyKey)) {
            return new ArrayList((Collection)this.queryListCache.get(propertyKey));
        } else {
            Object queryList;
            String basicPhaseNamePattern;
            if(!this.properties.containsKey(propertyKey)) {
                if(benchmarkPhase.isQueryOrderRandom()) {
                    if(!this.queryListCache.containsKey("shuffledQueryList")) {
                        basicPhaseNamePattern = queryOrderBasicPhase.getQueryListProperty(0);
                        if(!this.properties.containsKey(basicPhaseNamePattern)) {
                            throw new IllegalArgumentException("Property " + basicPhaseNamePattern + " is not defined, but is the basis for shuffling the query list.");
                        }

                        this.queryListCache.put("shuffledQueryList", this.getQueryList(queryOrderBasicPhase, 0));
                    }

                    queryList = (List)this.queryListCache.get("shuffledQueryList");
                    this.shuffleList((List)queryList, this.rnd);
                } else {
                    queryList = this.getQueryList(queryOrderBasicPhase, 0);
                }
            } else {
                queryList = new ArrayList();
                String[] var11;
                int var10 = (var11 = this.properties.getProperty(propertyKey).split(",")).length;

                label65:
                for(int var9 = 0; var9 < var10; ++var9) {
                    basicPhaseNamePattern = var11[var9];
                    String[] queryRange = basicPhaseNamePattern.trim().split("-");
                    switch(queryRange.length) {
                    case 1:
                        ((List)queryList).add(Integer.valueOf(Integer.parseInt(queryRange[0].trim())));
                        break;
                    case 2:
                        int startQuery = Integer.parseInt(queryRange[0]);
                        int endQuery = Integer.parseInt(queryRange[1]);
                        int i;
                        if(startQuery > endQuery) {
                            i = startQuery;

                            while(true) {
                                if(i < endQuery) {
                                    continue label65;
                                }

                                ((List)queryList).add(Integer.valueOf(i));
                                --i;
                            }
                        } else {
                            i = startQuery;

                            while(true) {
                                if(i > endQuery) {
                                    continue label65;
                                }

                                ((List)queryList).add(Integer.valueOf(i));
                                ++i;
                            }
                        }
                    default:
                        throw new IllegalArgumentException("Query numbers must be in the form X or X-Y, comma separated.");
                    }
                }
            }

            if(queryOrderCached) {
                this.queryListCache.put(propertyKey, new ArrayList((Collection)queryList));
            }

            return new ArrayList((Collection)queryList);
        }
    }

parseEnvironment()

parseEnvironment()读取系统的环境变量并解析。

Map env = System.getenv();
this.version = (String)env.get("BIG_BENCH_VERSION");
this.homeDir = (String)env.get("BIG_BENCH_HOME");
this.confDir = (String)env.get("BIG_BENCH_CONF_DIR");
this.runScript = (String)env.get("BIG_BENCH_BIN_DIR") + "/bigBench";
this.datagenDir = (String)env.get("BIG_BENCH_DATA_GENERATOR_DIR");
this.logDir = (String)env.get("BIG_BENCH_LOGS_DIR");
this.dataGenLogFile = (String)env.get("BIG_BENCH_DATAGEN_STAGE_LOG");
this.loadLogFile = (String)env.get("BIG_BENCH_LOADING_STAGE_LOG");
this.engine = (String)env.get("BIG_BENCH_ENGINE");
this.database = (String)env.get("BIG_BENCH_DATABASE");
this.mapTasks = (String)env.get("BIG_BENCH_MAP_TASKS");
this.numberOfParallelStreams = Integer.parseInt((String)env.get("BIG_BENCH_NUMBER_OF_PARALLEL_STREAMS"));
this.scaleFactor = Long.parseLong((String)env.get("BIG_BENCH_SCALE_FACTOR"));
this.stopAfterFailure = ((String)env.get("BIG_BENCH_STOP_AFTER_FAILURE")).equals("1");

并自动在用户指定的参数后面加上 -U (解锁专家模式)

this.userArguments.add("-U");

如果用户指定了 PRETEND_MODE、PRINT_STD_OUT、WORKLOAD、QUERIES_TO_RUN，则以用户指定的参数为准，否则使用默认值。

if(env.containsKey("USER_PRETEND_MODE")) {
    this.properties.setProperty("pretend_mode", (String)env.get("USER_PRETEND_MODE"));
}

if(env.containsKey("USER_PRINT_STD_OUT")) {
   this.properties.setProperty("show_command_stdout", (String)env.get("USER_PRINT_STD_OUT"));
}

if(env.containsKey("USER_DRIVER_WORKLOAD")) {
   this.properties.setProperty("workload", (String)env.get("USER_DRIVER_WORKLOAD"));
}

if(env.containsKey("USER_DRIVER_QUERIES_TO_RUN")) {
    this.properties.setProperty(BigBench.BenchmarkPhase.POWER_TEST.getQueryListProperty(0), (String)env.get("USER_DRIVER_QUERIES_TO_RUN"));
}

读取 workload 并赋值 benchmarkPhases。如果 workload 里不包含 BENCHMARK_START 和 BENCHMARK_STOP，自动在 benchmarkPhases 的首位和末位分别加上 BENCHMARK_START 和 BENCHMARK_STOP。

this.benchmarkPhases = new ArrayList();
Iterator var7 = Arrays.asList(this.properties.getProperty("workload").split(",")).iterator();

while(var7.hasNext()) {
    String benchmarkPhase = (String)var7.next();
    this.benchmarkPhases.add(BigBench.BenchmarkPhase.valueOf(benchmarkPhase.trim()));
}

if(!this.benchmarkPhases.contains(BigBench.BenchmarkPhase.BENCHMARK_START)) {
    this.benchmarkPhases.add(0, BigBench.BenchmarkPhase.BENCHMARK_START);
}

if(!this.benchmarkPhases.contains(BigBench.BenchmarkPhase.BENCHMARK_STOP)) {
    this.benchmarkPhases.add(BigBench.BenchmarkPhase.BENCHMARK_STOP);
}

run()

run() 方法是 RunBigBench.jar 里核心的方法。所有的执行都是通过 run() 方法调用的。比如 runQueries()、runModule()、generateData()等。runQueries()、runModule()、generateData() 又通过调用 runCmd() 方法来创建操作系统进程，执行bash命令，调用bash脚本。

run() 方法里通过一个 while 循环来逐一执行 workload 里的每一个 benchmarkPhase。 不同的 benchmarkPhase 会调用 runQueries()、runModule()、generateData()...中的不同方法。

try {
long e = 0L;
this.log.finer("Benchmark phases: " + this.benchmarkPhases);
Iterator startCheckpoint = this.benchmarkPhases.iterator();

long throughputStart;
while(startCheckpoint.hasNext()) {
    BigBench.BenchmarkPhase children = (BigBench.BenchmarkPhase)startCheckpoint.next();
    if(children.isPhaseDone()) {
        this.log.info("The phase " + children.name() + " was already performed earlier. Skipping this phase");
    } else {
        try {
            switch($SWITCH_TABLE$io$bigdatabenchmark$v1$driver$BigBench$BenchmarkPhase()[children.ordinal()]) {
            case 1:
            case 20:
                throw new IllegalArgumentException("The value " + children.name() + " is only used internally.");
            case 2:
                this.log.info(children.getConsoleMessage());
                e = System.currentTimeMillis();
                break;
            case 3:
                if(!BigBench.BenchmarkPhase.BENCHMARK_START.isPhaseDone()) {
                    throw new IllegalArgumentException("Error: Cannot stop the benchmark before starting it");
                }

                throughputStart = System.currentTimeMillis();
                this.log.info(String.format("%-55s finished. Time: %25s", new Object[]{children.getConsoleMessage(), BigBench.Helper.formatTime(throughputStart - e)}));
                this.logTreeRoot.setCheckpoint(new BigBench.Checkpoint(BigBench.BenchmarkPhase.BENCHMARK, -1L, -1L, e, throughputStart, this.logTreeRoot.isSuccessful()));
                break;
            case 4:
            case 15:
            case 18:
            case 22:
            case 27:
            case 28:
            case 29:
                this.runModule(children, this.userArguments);
                break;
            case 5:
            case 10:
            case 11:
                this.runQueries(children, 1, validationArguments);
                break;
            case 6:
            case 9:
                this.runModule(children, validationArguments);
                break;
            case 7:
                this.generateData(children, false, validationArguments);
                break;
            case 8:
                this.generateData(children, true, validationArguments);
                break;
            case 12:
            case 19:
            case 24:
                this.runQueries(children, 1, this.userArguments);
                break;
            case 13:
            case 14:
            case 21:
            case 23:
            case 25:
            case 26:
                this.runQueries(children, this.numberOfParallelStreams, this.userArguments);
                break;
            case 16:
                this.generateData(children, false, this.userArguments);
                break;
            case 17:
                this.generateData(children, true, this.userArguments);
            }

            children.setPhaseDone(true);
        } catch (IOException var21) {
            this.log.info("==============\nBenchmark run terminated\nReason: An error occured while running a command in phase " + children + "\n==============");
            var21.printStackTrace();
            if(this.stopAfterFailure || children.mustSucceed()) {
                break;
            }
        }
    }
}

这里的 case 1-29 并不是 1-29 条查询，而是枚举类型里的 1-29 个 benmarkPhase 。如下所示：

private static enum BenchmarkPhase {
BENCHMARK((String)null, "benchmark", false, false, false, false, "BigBench benchmark"),
BENCHMARK_START((String)null, "benchmark_start", false, false, false, false, "BigBench benchmark: Start"),
BENCHMARK_STOP((String)null, "benchmark_stop", false, false, false, false, "BigBench benchmark: Stop"),
CLEAN_ALL("cleanAll", "clean_all", false, false, false, false, "BigBench clean all"),
ENGINE_VALIDATION_CLEAN_POWER_TEST("cleanQuery", "engine_validation_power_test", false, false, false, false, "BigBench engine validation: Clean power test queries"),
ENGINE_VALIDATION_CLEAN_LOAD_TEST("cleanMetastore", "engine_validation_metastore", false, false, false, false, "BigBench engine validation: Clean metastore"),
ENGINE_VALIDATION_CLEAN_DATA("cleanData", "engine_validation_data", false, false, false, false, "BigBench engine validation: Clean data"),
ENGINE_VALIDATION_DATA_GENERATION("dataGen", "engine_validation_data", false, false, false, true, "BigBench engine validation: Data generation"),
ENGINE_VALIDATION_LOAD_TEST("populateMetastore", "engine_validation_metastore", false, false, false, true, "BigBench engine validation: Populate metastore"),
ENGINE_VALIDATION_POWER_TEST("runQuery", "engine_validation_power_test", false, false, false, false, "BigBench engine validation: Power test"),
ENGINE_VALIDATION_RESULT_VALIDATION("validateQuery", "engine_validation_power_test", false, false, true, false, "BigBench engine validation: Check all query results"),
CLEAN_POWER_TEST("cleanQuery", "power_test", false, false, false, false, "BigBench clean: Clean power test queries"),
CLEAN_THROUGHPUT_TEST_1("cleanQuery", "throughput_test_1", false, false, false, false, "BigBench clean: Clean first throughput test queries"),
CLEAN_THROUGHPUT_TEST_2("cleanQuery", "throughput_test_2", false, false, false, false, "BigBench clean: Clean second throughput test queries"),
CLEAN_LOAD_TEST("cleanMetastore", "metastore", false, false, false, false, "BigBench clean: Load test"),
CLEAN_DATA("cleanData", "data", false, false, false, false, "BigBench clean: Data"),
DATA_GENERATION("dataGen", "data", false, false, false, true, "BigBench preparation: Data generation"),
LOAD_TEST("populateMetastore", "metastore", false, false, false, true, "BigBench phase 1: Load test"),
POWER_TEST("runQuery", "power_test", false, true, false, false, "BigBench phase 2: Power test"),
THROUGHPUT_TEST((String)null, "throughput_test", false, false, false, false, "BigBench phase 3: Throughput test"),
THROUGHPUT_TEST_1("runQuery", "throughput_test_1", true, true, false, false, "BigBench phase 3: First throughput test run"),
THROUGHPUT_TEST_REFRESH("refreshMetastore", "throughput_test_refresh", false, false, false, false, "BigBench phase 3: Throughput test data refresh"),
THROUGHPUT_TEST_2("runQuery", "throughput_test_2", true, true, false, false, "BigBench phase 3: Second throughput test run"),
VALIDATE_POWER_TEST("validateQuery", "power_test", false, false, true, false, "BigBench validation: Power test results"),
VALIDATE_THROUGHPUT_TEST_1("validateQuery", "throughput_test_1", false, false, true, false, "BigBench validation: First throughput test results"),
VALIDATE_THROUGHPUT_TEST_2("validateQuery", "throughput_test_2", false, false, true, false, "BigBench validation: Second throughput test results"),
SHOW_TIMES("showTimes", "show_times", false, false, true, false, "BigBench: show query times"),
SHOW_ERRORS("showErrors", "show_errors", false, false, true, false, "BigBench: show query errors"),
SHOW_VALIDATION("showValidation", "show_validation", false, false, true, false, "BigBench: show query validation results");

private String runModule;
private String namePattern;
private boolean queryOrderRandom;
private boolean queryOrderCached;
private boolean printStdOut;
private boolean mustSucceed;
private String consoleMessage;
private boolean phaseDone;

private BenchmarkPhase(String runModule, String namePattern, boolean queryOrderRandom, boolean queryOrderCached, boolean printStdOut, boolean mustSucceed, String consoleMessage) {
    this.runModule = runModule;
    this.namePattern = namePattern;
    this.queryOrderRandom = queryOrderRandom;
    this.queryOrderCached = queryOrderCached;
    this.printStdOut = printStdOut;
    this.mustSucceed = mustSucceed;
    this.consoleMessage = consoleMessage;
    this.phaseDone = false;
}

3对应 BENCHMARK_STOP，4对应 CLEAN_ALL,29对应 SHOW_VALIDATION，依此类推...

可以看出：

CLEAN_ALL、CLEAN_LOAD_TEST、LOAD_TEST、THROUGHPUT_TEST_REFRESH、SHOW_TIMES、SHOW_ERRORS、SHOW_VALIDATION等benchmarkPhases调用的是

this.runModule(children, this.userArguments);

方法是 runModule ，参数是 this.userArguments。

ENGINE_VALIDATION_CLEAN_POWER_TEST、ENGINE_VALIDATION_POWER_TEST、ENGINE_VALIDATION_RESULT_VALIDATION 调用的是

this.runQueries(children, 1, validationArguments);

方法是 runQueries ，参数是 1(stream number) 和 validationArguments。

ENGINE_VALIDATION_CLEAN_LOAD_TEST 和 ENGINE_VALIDATION_LOAD_TEST 调用的是

this.runModule(children, validationArguments);

ENGINE_VALIDATION_CLEAN_DATA 调用的是

this.generateData(children, false, validationArguments);

ENGINE_VALIDATION_DATA_GENERATION 调用的是

this.generateData(children, true, validationArguments);

CLEAN_POWER_TEST、POWER_TEST、VALIDATE_POWER_TEST 调用的是

this.runQueries(children, 1, this.userArguments);

CLEAN_THROUGHPUT_TEST_1``CLEAN_THROUGHPUT_TEST_2``THROUGHPUT_TEST_1``THROUGHPUT_TEST_2``VALIDATE_THROUGHPUT_TEST_1 VALIDATE_THROUGHPUT_TEST_2 调用的是

this.runQueries(children, this.numberOfParallelStreams, this.userArguments);

CLEAN_DATA 调用的是

this.generateData(children, false, this.userArguments);

DATA_GENERATION 调用的是

this.generateData(children, true, this.userArguments);

总结一下以上的方法调用可以发现：

  • 跟 ENGINE_VALIDATION 相关的benchmarkPhase用的参数都是 validationArguments。其余用的是 userArguments（ validationArguments 和 userArguments 唯一的区别是 validationArguments 的 SCALE_FACTOR 恒为1）
  • 跟 POWER_TEST 相关的都是调用 runQueries() 方法，因为 POWER_TEST 就是执行SQL查询
  • 跟 CLEAN_DATA DATA_GENERATION 相关的都是调用 generateData() 方法
  • 跟 LOAD_TEST SHOW 相关的都是调用 runModule() 方法

benchmarkPhase 和 module 对应关系

具体每个 benchmarkPhase 跟 module（执行的脚本）的对应关系如下：

CLEAN_ALL -> "cleanAll"
ENGINE_VALIDATION_CLEAN_POWER_TEST -> "cleanQuery"
ENGINE_VALIDATION_CLEAN_LOAD_TEST -> "cleanMetastore",
ENGINE_VALIDATION_CLEAN_DATA -> "cleanData"
ENGINE_VALIDATION_DATA_GENERATION -> "dataGen"
ENGINE_VALIDATION_LOAD_TEST -> "populateMetastore"
ENGINE_VALIDATION_POWER_TEST -> "runQuery"
ENGINE_VALIDATION_RESULT_VALIDATION -> "validateQuery"
CLEAN_POWER_TEST -> "cleanQuery"
CLEAN_THROUGHPUT_TEST_1 -> "cleanQuery"
CLEAN_THROUGHPUT_TEST_2 -> "cleanQuery"
CLEAN_LOAD_TEST -> "cleanMetastore"
CLEAN_DATA -> "cleanData"
DATA_GENERATION -> "dataGen"
LOAD_TEST -> "populateMetastore"
POWER_TEST -> "runQuery"
THROUGHPUT_TEST -> (String)null
THROUGHPUT_TEST_1 -> "runQuery"
THROUGHPUT_TEST_REFRESH -> "refreshMetastore"
THROUGHPUT_TEST_2 -> "runQuery"
VALIDATE_POWER_TEST -> "validateQuery"
VALIDATE_THROUGHPUT_TEST_1 -> "validateQuery"
VALIDATE_THROUGHPUT_TEST_2 -> "validateQuery"
SHOW_TIMES -> "showTimes"
SHOW_ERRORS -> "showErrors"
SHOW_VALIDATION -> "showValidation"

当执行某个 benchmarkPhase 时会去调用如上该 benchmarkPhase 对应的 module （脚本位于 $BENCH_MARK_HOME/engines/hive/bin 目录下）

cmdLine.add(benchmarkPhase.getRunModule());

程序调用流程

bigBench.png

接下来介绍每个module的功能

module

cleanAll

1. DROP DATABASE
2. 删除hdfs上的源数据
echo "dropping database (with all tables)"
runCmdWithErrorCheck runEngineCmd -e "DROP DATABASE IF EXISTS $BIG_BENCH_DATABASE CASCADE;"
echo "cleaning ${BIG_BENCH_HDFS_ABSOLUTE_HOME}"
hadoop fs -rm -r -f -skipTrash "${BIG_BENCH_HDFS_ABSOLUTE_HOME}"

cleanQuery

1. 删除对应的 Query 生成的临时表
2. 删除对应的 Query 生成的结果表
runCmdWithErrorCheck runEngineCmd -e "DROP TABLE IF EXISTS $TEMP_TABLE1; DROP TABLE IF EXISTS $TEMP_TABLE2; DROP TABLE IF EXISTS $RESULT_TABLE;"
return $?

cleanMetastore

1. 调用 `dropTables.sql` 将23张表依次DROP
echo "cleaning metastore tables"
runCmdWithErrorCheck runEngineCmd -f "$BIG_BENCH_CLEAN_METASTORE_FILE"
export BIG_BENCH_CLEAN_METASTORE_FILE="$BIG_BENCH_CLEAN_DIR/dropTables.sql"

dropTables.sql 将23张表依次DROP,源码如下：

DROP TABLE IF EXISTS ${hiveconf:customerTableName};
DROP TABLE IF EXISTS ${hiveconf:customerAddressTableName};
DROP TABLE IF EXISTS ${hiveconf:customerDemographicsTableName};
DROP TABLE IF EXISTS ${hiveconf:dateTableName};
DROP TABLE IF EXISTS ${hiveconf:householdDemographicsTableName};
DROP TABLE IF EXISTS ${hiveconf:incomeTableName};
DROP TABLE IF EXISTS ${hiveconf:itemTableName};
DROP TABLE IF EXISTS ${hiveconf:promotionTableName};
DROP TABLE IF EXISTS ${hiveconf:reasonTableName};
DROP TABLE IF EXISTS ${hiveconf:shipModeTableName};
DROP TABLE IF EXISTS ${hiveconf:storeTableName};
DROP TABLE IF EXISTS ${hiveconf:timeTableName};
DROP TABLE IF EXISTS ${hiveconf:warehouseTableName};
DROP TABLE IF EXISTS ${hiveconf:webSiteTableName};
DROP TABLE IF EXISTS ${hiveconf:webPageTableName};
DROP TABLE IF EXISTS ${hiveconf:inventoryTableName};
DROP TABLE IF EXISTS ${hiveconf:storeSalesTableName};
DROP TABLE IF EXISTS ${hiveconf:storeReturnsTableName};
DROP TABLE IF EXISTS ${hiveconf:webSalesTableName};
DROP TABLE IF EXISTS ${hiveconf:webReturnsTableName};
DROP TABLE IF EXISTS ${hiveconf:marketPricesTableName};
DROP TABLE IF EXISTS ${hiveconf:clickstreamsTableName};
DROP TABLE IF EXISTS ${hiveconf:reviewsTableName};

cleanData

1. 删除hdfs上 /user/root/benchmarks/bigbench/data 目录
2. 删除hdfs上 /user/root/benchmarks/bigbench/data_refresh 目录
echo "cleaning ${BIG_BENCH_HDFS_ABSOLUTE_INIT_DATA_DIR}"
hadoop fs -rm -r -f -skipTrash "${BIG_BENCH_HDFS_ABSOLUTE_INIT_DATA_DIR}"
echo "cleaning ${BIG_BENCH_HDFS_ABSOLUTE_REFRESH_DATA_DIR}"
hadoop fs -rm -r -f -skipTrash "${BIG_BENCH_HDFS_ABSOLUTE_REFRESH_DATA_DIR}"

dataGen

1. 创建目录 /user/root/benchmarks/bigbench/data 并赋予权限
2. 创建目录 /user/root/benchmarks/bigbench/data_refresh 并赋予权限
3. 调用 HadoopClusterExec.jar 和 pdgf.jar 生成 base data 到 /user/root/benchmarks/bigbench/data 目录下
4. 调用 HadoopClusterExec.jar 和 pdgf.jar 生成 refresh data 到 /user/root/benchmarks/bigbench/data_refresh 目录下

创建目录 /user/root/benchmarks/bigbench/data 并赋予权限

runCmdWithErrorCheck hadoop fs -mkdir -p "${BIG_BENCH_HDFS_ABSOLUTE_INIT_DATA_DIR}"
runCmdWithErrorCheck hadoop fs -chmod 777 "${BIG_BENCH_HDFS_ABSOLUTE_INIT_DATA_DIR}"

创建目录 /user/root/benchmarks/bigbench/data_refresh 并赋予权限

runCmdWithErrorCheck hadoop fs -mkdir -p "${BIG_BENCH_HDFS_ABSOLUTE_REFRESH_DATA_DIR}"
runCmdWithErrorCheck hadoop fs -chmod 777 "${BIG_BENCH_HDFS_ABSOLUTE_REFRESH_DATA_DIR}"

调用 HadoopClusterExec.jar 和 pdgf.jar 生成 base data

runCmdWithErrorCheck hadoop jar "${BIG_BENCH_TOOLS_DIR}/HadoopClusterExec.jar" -archives "${PDGF_ARCHIVE_PATH}" ${BIG_BENCH_DATAGEN_HADOOP_EXEC_DEBUG} -taskFailOnNonZeroReturnValue -execCWD "${PDGF_DISTRIBUTED_NODE_DIR}" ${HadoopClusterExecOptions} -exec ${BIG_BENCH_DATAGEN_HADOOP_JVM_ENV} -cp "${HADOOP_CP}:pdgf.jar" ${PDGF_CLUSTER_CONF} pdgf.Controller -nc HadoopClusterExec.tasks -nn HadoopClusterExec.taskNumber -ns -c -sp REFRESH_PHASE 0 -o "‘${BIG_BENCH_HDFS_ABSOLUTE_INIT_DATA_DIR}/‘+table.getName()+‘/‘" ${BIG_BENCH_DATAGEN_HADOOP_OPTIONS} -s ${BIG_BENCH_DATAGEN_TABLES} ${PDGF_OPTIONS} "[email protected]" 2>&1 | tee -a "$BIG_BENCH_DATAGEN_STAGE_LOG" 2>&1

调用 HadoopClusterExec.jar 和 pdgf.jar 生成 refresh data

runCmdWithErrorCheck hadoop jar "${BIG_BENCH_TOOLS_DIR}/HadoopClusterExec.jar" -archives "${PDGF_ARCHIVE_PATH}" ${BIG_BENCH_DATAGEN_HADOOP_EXEC_DEBUG} -taskFailOnNonZeroReturnValue -execCWD "${PDGF_DISTRIBUTED_NODE_DIR}" ${HadoopClusterExecOptions} -exec ${BIG_BENCH_DATAGEN_HADOOP_JVM_ENV} -cp "${HADOOP_CP}:pdgf.jar" ${PDGF_CLUSTER_CONF} pdgf.Controller -nc HadoopClusterExec.tasks -nn HadoopClusterExec.taskNumber -ns -c -sp REFRESH_PHASE 1 -o "‘${BIG_BENCH_HDFS_ABSOLUTE_REFRESH_DATA_DIR}/‘+table.getName()+‘/‘" ${BIG_BENCH_DATAGEN_HADOOP_OPTIONS} -s ${BIG_BENCH_DATAGEN_TABLES} ${PDGF_OPTIONS} "[email protected]" 2>&1 | tee -a "$BIG_BENCH_DATAGEN_STAGE_LOG" 2>&1

populateMetastore

该过程是真正的创建数据库表的过程。建表的过程调用的是 $BENCH_MARK_HOME/engines/hive/population/ 目录下的 hiveCreateLoad.sql ,通过该sql文件来建数据库表。

  1. 从 /user/root/benchmarks/bigbench/data 路径下读取 .dat 的原始数据，生成 TEXTFILE 格式的外部临时表
  2. 用 select * from 临时表 来创建最终的 ORC 格式的数据库表
  3. 删除外部临时表。

从 /user/root/benchmarks/bigbench/data 路径下读取 .dat 的原始数据，生成 TEXTFILE 格式的外部临时表

DROP TABLE IF EXISTS ${hiveconf:customerTableName}${hiveconf:temporaryTableSuffix};
CREATE EXTERNAL TABLE ${hiveconf:customerTableName}${hiveconf:temporaryTableSuffix}
  ( c_customer_sk             bigint              --not null
  , c_customer_id             string              --not null
  , c_current_cdemo_sk        bigint
  , c_current_hdemo_sk        bigint
  , c_current_addr_sk         bigint
  , c_first_shipto_date_sk    bigint
  , c_first_sales_date_sk     bigint
  , c_salutation              string
  , c_first_name              string
  , c_last_name               string
  , c_preferred_cust_flag     string
  , c_birth_day               int
  , c_birth_month             int
  , c_birth_year              int
  , c_birth_country           string
  , c_login                   string
  , c_email_address           string
  , c_last_review_date        string
  )
  ROW FORMAT DELIMITED FIELDS TERMINATED BY ‘${hiveconf:fieldDelimiter}‘
  STORED AS TEXTFILE LOCATION ‘${hiveconf:hdfsDataPath}/${hiveconf:customerTableName}‘
;

用 select * from 临时表 来创建最终的 ORC 格式的数据库表

DROP TABLE IF EXISTS ${hiveconf:customerTableName};
CREATE TABLE ${hiveconf:customerTableName}
STORED AS ${hiveconf:tableFormat}
AS
SELECT * FROM ${hiveconf:customerTableName}${hiveconf:temporaryTableSuffix}
;

删除外部临时表

DROP TABLE ${hiveconf:customerTableName}${hiveconf:temporaryTableSuffix};

runQuery

1. runQuery 调用每个query下的 run.sh 里的 `query_run_main_method()` 方法
2. `query_run_main_method()` 调用 `runEngineCmd` 来执行query脚本（qxx.sql）

runQuery 调用每个query下的 run.sh 里的 query_run_main_method() 方法

QUERY_MAIN_METHOD="query_run_main_method"
-----------------------------------------
"$QUERY_MAIN_METHOD" 2>&1 | tee -a "$LOG_FILE_NAME" 2>&1

query_run_main_method() 调用 runEngineCmd 来执行query脚本（qxx.sql）

query_run_main_method () {
    QUERY_SCRIPT="$QUERY_DIR/$QUERY_NAME.sql"
    if [ ! -r "$QUERY_SCRIPT" ]
    then
        echo "SQL file $QUERY_SCRIPT can not be read."
        exit 1
    fi

    runCmdWithErrorCheck runEngineCmd -f "$QUERY_SCRIPT"
    return $?
}

一般情况下 query_run_main_method () 方法只是执行对应的query脚本，但是像 q05、q20... 这些查询，用到了机器学习算法，所以在执行对应的query脚本后会把生成的结果表作为输入，然后调用执行机器学习算法（如聚类、逻辑回归）的jar包继续执行，得到最终的结果。

runEngineCmd () {
  if addInitScriptsToParams
  then
    "$BINARY" "${BINARY_PARAMS[@]}" "${INIT_PARAMS[@]}" "[email protected]"
  else
    return 1
  fi
}
--------------------------
BINARY="/usr/bin/hive"
BINARY_PARAMS+=(--hiveconf BENCHMARK_PHASE=$BIG_BENCH_BENCHMARK_PHASE --hiveconf STREAM_NUMBER=$BIG_BENCH_STREAM_NUMBER --hiveconf QUERY_NAME=$QUERY_NAME --hiveconf QUERY_DIR=$QUERY_DIR --hiveconf RESULT_TABLE=$RESULT_TABLE --hiveconf RESULT_DIR=$RESULT_DIR --hiveconf TEMP_TABLE=$TEMP_TABLE --hiveconf TEMP_DIR=$TEMP_DIR --hiveconf TABLE_PREFIX=$TABLE_PREFIX)
INIT_PARAMS=(-i "$BIG_BENCH_QUERY_PARAMS_FILE" -i "$BIG_BENCH_ENGINE_SETTINGS_FILE")
INIT_PARAMS+=(-i "$LOCAL_QUERY_ENGINE_SETTINGS_FILE")

if [ -n "$USER_QUERY_PARAMS_FILE" ]
then
if [ -r "$USER_QUERY_PARAMS_FILE" ]
then
  echo "User defined query parameter file found. Adding $USER_QUERY_PARAMS_FILE to hive init."
  INIT_PARAMS+=(-i "$USER_QUERY_PARAMS_FILE")
else
  echo "User query parameter file $USER_QUERY_PARAMS_FILE can not be read."
  return 1
fi
fi

if [ -n "$USER_ENGINE_SETTINGS_FILE" ]
then
if [ -r "$USER_ENGINE_SETTINGS_FILE" ]
then
  echo "User defined engine settings file found. Adding $USER_ENGINE_SETTINGS_FILE to hive init."
  INIT_PARAMS+=(-i "$USER_ENGINE_SETTINGS_FILE")
else
  echo "User hive settings file $USER_ENGINE_SETTINGS_FILE can not be read."
  return 1
fi
fi
return 0

validateQuery

1. 调用每个query下的 run.sh 里的 `query_run_validate_method()` 方法
2. `query_run_validate_method()` 比较 `$BENCH_MARK_HOME/engines/hive/queries/qxx/results/qxx-result` 和hdfs上 `/user/root/benchmarks/bigbench/queryResults/qxx_hive_${BIG_BENCH_BENCHMARK_PHASE}_${BIG_BENCH_STREAM_NUMBER}_result` 两个文件，如果一样，则验证通过，否则验证失败。
if diff -q "$VALIDATION_RESULTS_FILENAME" <(hadoop fs -cat "$RESULT_DIR/*")
then
    echo "Validation of $VALIDATION_RESULTS_FILENAME passed: Query returned correct results"
else
    echo "Validation of $VALIDATION_RESULTS_FILENAME failed: Query returned incorrect results"
    VALIDATION_PASSED="0"
fi

SF为1时(-f 1)，用上面的方法比较，SF不为1（>1）时,只要hdfs上的结果表中行数大于等于1即验证通过

if [ `hadoop fs -cat "$RESULT_DIR/*" | head -n 10 | wc -l` -ge 1 ]
then
    echo "Validation passed: Query returned results"
else
    echo "Validation failed: Query did not return results"
    return 1
fi

refreshMetastore

1. 调用 `$BENCH_MARK_HOME/engines/hive/refresh/` 目录下的 `hiveRefreshCreateLoad.sql` 脚本
2. `hiveRefreshCreateLoad.sql` 将hdfs上 `/user/root/benchmarks/bigbench/data_refresh/` 目录下每个表数据插入外部临时表
3. 外部临时表再将每个表的数据插入Hive数据库对应的表中

hiveRefreshCreateLoad.sql 将hdfs上 /user/root/benchmarks/bigbench/data_refresh/ 目录下每个表数据插入外部临时表

DROP TABLE IF EXISTS ${hiveconf:customerTableName}${hiveconf:temporaryTableSuffix};
CREATE EXTERNAL TABLE ${hiveconf:customerTableName}${hiveconf:temporaryTableSuffix}
  ( c_customer_sk             bigint              --not null
  , c_customer_id             string              --not null
  , c_current_cdemo_sk        bigint
  , c_current_hdemo_sk        bigint
  , c_current_addr_sk         bigint
  , c_first_shipto_date_sk    bigint
  , c_first_sales_date_sk     bigint
  , c_salutation              string
  , c_first_name              string
  , c_last_name               string
  , c_preferred_cust_flag     string
  , c_birth_day               int
  , c_birth_month             int
  , c_birth_year              int
  , c_birth_country           string
  , c_login                   string
  , c_email_address           string
  , c_last_review_date        string
  )
  ROW FORMAT DELIMITED FIELDS TERMINATED BY ‘${hiveconf:fieldDelimiter}‘
  STORED AS TEXTFILE LOCATION ‘${hiveconf:hdfsDataPath}/${hiveconf:customerTableName}‘
;
-----------------
set hdfsDataPath=${env:BIG_BENCH_HDFS_ABSOLUTE_REFRESH_DATA_DIR};

外部临时表再将每个表的数据插入Hive数据库对应的表中

INSERT INTO TABLE ${hiveconf:customerTableName}
SELECT * FROM ${hiveconf:customerTableName}${hiveconf:temporaryTableSuffix}
;

附录

23张数据库表

时间： 2024-11-24 16:45:17

TPCx-BB源码分析的相关文章

【Cocos2d-x】源码分析之 2d/ui/Widget

从今天开始 咱也模仿 红孩儿这些大牛分析源码 ,由于水平有限 不对之处欢迎狂喷.哈哈. #ifndef __UIWIDGET_H__ #define __UIWIDGET_H__ #include "ui/CCProtectedNode.h" #include "ui/UILayoutDefine.h" #include "ui/UILayoutParameter.h" #include "ui/GUIDefine.h" NS

【Zookeeper】源码分析之持久化--FileTxnSnapLog

一.前言 前面分析了FileSnap,接着继续分析FileTxnSnapLog源码,其封装了TxnLog和SnapShot,其在持久化过程中是一个帮助类. 二.FileTxnSnapLog源码分析 2.1 类的属性 public class FileTxnSnapLog { //the direcotry containing the //the transaction logs // 日志文件目录 private final File dataDir; //the directory cont

cocos2d-x 源码分析 之 CCTableView源码分析（附使用方法讨论）

cocos2d-x源码总目录 http://blog.csdn.net/u011225840/article/details/31743129 源码来自2.x,转载请注明 1.继承结构 首先来看下CCTableView的继承结构 从继承结构上看,CCTableView是一种CCScrollView,所以为了研究CCTableView的源码,清先去了解CCScrollView的源码http://blog.csdn.net/u011225840/article/details/30033501. 其

【Zookeeper】源码分析之持久化--FileSnap

一.前言 前篇博文已经分析了FileTxnLog的源码,现在接着分析持久化中的FileSnap,其主要提供了快照相应的接口. 二.SnapShot源码分析 SnapShot是FileTxnLog的父类,接口类型,其方法如下 public interface SnapShot { /** * deserialize a data tree from the last valid snapshot and * return the last zxid that was deserialized *

JDK1.8源码分析之TreeMap（五）

一.前言 当我们需要把插入的元素进行排序的时候,就是时候考虑TreeMap了,从名字上来看,TreeMap肯定是和树是脱不了干系的,它是一个排序了的Map,下面我们来着重分析其源码,理解其底层如何实现排序功能.下面,开始分析. 二.TreeMap示例 import java.util.TreeMap; import java.util.Map; public class TreeMapTest { public static void main(String[] args) { Map<Stri

JDK1.8源码分析之LinkedHashMap（二）

一.前言 前面我们已经分析了HashMap的源码,已经知道了HashMap可以用在哪种场合,如果这样一种情形,我们需要按照元素插入的顺序来访问元素,此时,LinkedHashMap就派上用场了,它保存着元素插入的顺序,并且可以按照我们插入的顺序进行访问. 二.LinkedHashMap用法 import java.util.Map; import java.util.LinkedHashMap; public class Test { public static void main(String

zg手册 之 python2.7.7源码分析（2）-- python 的整数对象和字符串对象

python 中的内置对象 python 中常用的内置对象有:整数对象,字符串对象,列表对象,字典对象.这些对象在python中使用最多,所以在实现上提供缓存机制,以提高运行效率. 整数对象 (PyIntObject) python 中的整数对象是不可变对象(immutable),即创建了一个 python 整数对象之后,不能再改变该对象的值. python 为创建整数对象提供了下面三种方法,其中 PyInt_FromString 和 PyInt_FromUnicode 内部也是调用 PyInt

zookeeper源码分析之一服务端处理请求流程

上文: zookeeper源码分析之一服务端启动过程 中,我们介绍了zookeeper服务器的启动过程,其中单机是ZookeeperServer启动,集群使用QuorumPeer启动,那么这次我们分析各自一下消息处理过程: 前文可以看到在 1.在单机情况下NettyServerCnxnFactory中启动ZookeeperServer来处理消息: public synchronized void startup() { if (sessionTracker == null) { createSe

Zepto源码分析之二~三个API

由于时间关系:本次只对这三个API($.camelCase.$.contains.$.each)方法进行分析 第一个方法变量转驼峰:$.camelCase('hello-world-welcome'); 源码: var camelize; /** * 字符串替换 * 使用replace第二个参数带回调 */ camelize = function(str) { return str.replace(/-+(.)?/g, function(match, chr) { return chr ? ch

Docker源码分析(二)：Docker Client创建与命令执行

1. 前言 如今,Docker作为业界领先的轻量级虚拟化容器管理引擎,给全球开发者提供了一种新颖.便捷的软件集成测试与部署之道.在团队开发软件时,Docker可以提供可复用的运行环境.灵活的资源配置.便捷的集成测试方法以及一键式的部署方式.可以说,Docker的优势在简化持续集成.运维部署方面体现得淋漓尽致,它完全让开发者从持续集成.运维部署方面中解放出来,把精力真正地倾注在开发上. 然而,把Docker的功能发挥到极致,并非一件易事.在深刻理解Docker架构的情况下,熟练掌握Docker C