neoremind / 2018-polar-race

1st AliCloud Database Performance Competition in 2018 - Java rank No.1 source code 阿里云2018年第一届PolarDB数据库性能大赛Java排名第一源码

天池大赛-阿里云2018年第一届PolarDB数据库性能大赛分享

天池比赛链接

赛题描述

解题报告长文

How to build

cd engine_java && mvn assembly:assembly

How to run

默认只能在Linux上运行（因为使用了JNA的Direct I/O）。

Write:

java -server -XX:-UseBiasedLocking -Xms2000m -Xmx2000m -XX:NewSize=1400m -XX:MaxMetaspaceSize=32m -XX:MaxDirectMemorySize=1G -XX:+UseG1GC -cp engine_java-0.0.1-SNAPSHOT-jar-with-dependencies.jar com.alibabacloud.polar_race.engine.common.neoremind.benchmark.BenchMark2 64 1000000 8 4096 /please/set/directory 1

Read:

java -server -XX:-UseBiasedLocking -Xms2000m -Xmx2000m -XX:NewSize=1400m -XX:MaxMetaspaceSize=32m -XX:MaxDirectMemorySize=1G -XX:+UseG1GC -cp engine_java-0.0.1-SNAPSHOT-jar-with-dependencies.jar com.alibabacloud.polar_race.engine.common.neoremind.benchmark.BenchMark2 64 1000000 8 4096 /please/set/directory 2

Range:

java -server -XX:-UseBiasedLocking -Xms2000m -Xmx2000m -XX:NewSize=1400m -XX:MaxMetaspaceSize=32m -XX:MaxDirectMemorySize=1G -XX:+UseG1GC -cp engine_java-0.0.1-SNAPSHOT-jar-with-dependencies.jar com.alibabacloud.polar_race.engine.common.neoremind.benchmark.BenchMark2 64 1000000 8 4096 /please/set/directory 2

Note:

64 1000000 8 4096 表示64并发，每个线程100w kv，key 8B，value 4K

2018-polar-race's Issues

增加远程debug

java -server -XX:-UseBiasedLocking -Xms2000m -Xmx2000m -XX:NewSize=1400m -XX:MaxMetaspaceSize=32m -XX:MaxDirectMemorySize=1G -XX:+UseG1GC -agentlib:jdwp=transport=dt_socket,server=y,suspend=n,address=8008 -cp engine_java-0.0.1-SNAPSHOT-jar-with-dependencies.jar com.alibabacloud.polar_race.engine.common.neoremind.benchmark.BenchMark2 16 1000 8 4096 /root/app/engine_java/please/set/directory 1

pom.xml文件中替换maven-assembly-plugin部门。原因，使用原代码打成的jar 不包含第三方的jar文件，运行是报错

pom.xml文件中替换maven-assembly-plugin部门。原因，使用原代码打成的jar 不包含第三方的jar文件，运行是报错：ClassNotFound:xxxx

<plugin>
  <groupId>org.apache.maven.plugins</groupId>
  <artifactId>maven-assembly-plugin</artifactId>
  <version>2.4</version>
  <configuration>
      <descriptorRefs>
          <descriptorRef>jar-with-dependencies</descriptorRef>
      </descriptorRefs>
  </configuration>
  <executions>
      <execution>
          <phase>package</phase>
          <goals>
              <goal>single</goal>
          </goals>
      </execution>
  </executions>
</plugin>

解题报告

您好，解题报告的链接挂掉了，请问有新的链接吗？

RangePartitioner.partition的计算逻辑

public class RangePartitioner implements Partitioner {

public RangePartitioner(int numberOfPartition) {
checkArgument(numberOfPartition == 1024, "partition number only support 1024 now");
}

/**

• // TODO assume partition number is 1024
  */
  @OverRide
  public int partition(byte[] key) {
  int result = ((0x03 & (key[0] >>> 6))) << 8;
  int subResult = ((0x3F & key[0]) << 2);
  int lastResult = 0x03 & (key[1] >>> 6);
  return result + subResult + lastResult;
  }
  }

RangePartitioner.partition，这个方法的逻辑是啥意思，没弄懂，求指教