AI 分析 MapReduce 之 Shuffle

ShuffleHandler

hadoop-mapreduce-project/hadoop-mapreduce-client/hadoop-mapreduce-client-shuffle/src/main/java/org/apache/hadoop/mapred/ShuffleHandler.java

cpp 复制代码
      "mapreduce_shuffle";

  public static final String SHUFFLE_PORT_CONFIG_KEY = "mapreduce.shuffle.port";
  public static final int DEFAULT_SHUFFLE_PORT = 13562;

  public static final String SHUFFLE_LISTEN_QUEUE_SIZE =
      "mapreduce.shuffle.listen.queue.size";
  public static final int DEFAULT_SHUFFLE_LISTEN_QUEUE_SIZE = 128;

  public static final String SHUFFLE_CONNECTION_KEEP_ALIVE_ENABLED =
      "mapreduce.shuffle.connection-keep-alive.enable";
  public static final boolean DEFAULT_SHUFFLE_CONNECTION_KEEP_ALIVE_ENABLED = false;

  public static final String SHUFFLE_CONNECTION_KEEP_ALIVE_TIME_OUT =
      "mapreduce.shuffle.connection-keep-alive.timeout";
  public static final int DEFAULT_SHUFFLE_CONNECTION_KEEP_ALIVE_TIME_OUT = 5; //seconds

  public static final String SHUFFLE_MAPOUTPUT_META_INFO_CACHE_SIZE =
      "mapreduce.shuffle.mapoutput-info.meta.cache.size";
  public static final int DEFAULT_SHUFFLE_MAPOUTPUT_META_INFO_CACHE_SIZE =
      1000;

  public static final String CONNECTION_CLOSE = "close";

  public static final String SUFFLE_SSL_FILE_BUFFER_SIZE_KEY =
    "mapreduce.shuffle.ssl.file.buffer.size";

  public static final int DEFAULT_SUFFLE_SSL_FILE_BUFFER_SIZE = 60 * 1024;

  public static final String MAX_SHUFFLE_CONNECTIONS = "mapreduce.shuffle.max.connections";
  public static final int DEFAULT_MAX_SHUFFLE_CONNECTIONS = 0; // 0 implies no limit
  
  public static final String MAX_SHUFFLE_THREADS = "mapreduce.shuffle.max.threads";
  // 0 implies Netty default of 2 * number of available processors
  public static final int DEFAULT_MAX_SHUFFLE_THREADS = 0;
  
  public static final String SHUFFLE_BUFFER_SIZE = 
      "mapreduce.shuffle.transfer.buffer.size";
  public static final int DEFAULT_SHUFFLE_BUFFER_SIZE = 128 * 1024;
  
  public static final String  SHUFFLE_TRANSFERTO_ALLOWED = 
      "mapreduce.shuffle.transferTo.allowed";
  public static final boolean DEFAULT_SHUFFLE_TRANSFERTO_ALLOWED = true;
  public static final boolean WINDOWS_DEFAULT_SHUFFLE_TRANSFERTO_ALLOWED = 
      false;
  private static final String TIMEOUT_HANDLER = "timeout";

  /* the maximum number of files a single GET request can
   open simultaneously during shuffle
   */
相关推荐
珠海西格电力1 小时前
零碳园区管理系统的性能评估体系的行业标准
大数据·网络·人工智能·物联网·机器学习·架构·能源
杨先生哦1 小时前
SOC告警研判实战:海量告警降噪、误报甄别、内网风险处置全流程(上岗必用)
大数据
一线数智1 小时前
朱鹰:人的判断力,比算力更重要
大数据·人工智能
用户3610588626122 小时前
Flink ResourceManager启动流程源码深度剖析:从ClusterEntrypoint到Slot分配的完整链路
大数据
用户3610588626122 小时前
Flink ApplicationMaster启动流程源码深度剖析:从YARN提交到JobMaster启动的完整链路
大数据·flink
初心未改HD3 小时前
DSH Skill开发实战:SKILL.md是给LLM看的操作手册,不是给人看的README,踩了4个坑
java·大数据·开发语言
程序 代码狂人3 小时前
据仓库常见表类型详解(面试完整版,金融数仓适用)
大数据
阿里云大数据AI技术4 小时前
云栖实录|Hologres 5.0全新发布:Agent Native 的全模态实时数仓
大数据·人工智能·agent
小天源4 小时前
品牌 AI 智能监测与自动化采集系统
大数据·人工智能·自动化
SelectDB5 小时前
从实时入库到实时可用:Apache Doris 5.0 增量计算场景与实践
大数据·数据库·开源