gfdsdg

分区排序溢写为小文件

hadoop1 reduce1spack1 reduce1

80%以后则溢写

spack1 reduce2word1 reduce2word1 reduce2

Map task1

word1 reduce2

调用Reduce方法

分为多个小文件

spack1 reduce1spacke1 reduce2word1 reduce2

合并为一个大文件

hadoop1 reduce1hello1 reduce1spack1 reduce1word1 reduce2

spack{1，1}word{1，1 }

分组

spack1 reduce1word1 reduce2

Reduce端shuffle开始

spack1 reduce2word1 reduce2

Reduce task1

hadoop {1，1}hello{1，1} spack{1，1}

hadoop1 reduce1hadoop1 reduce1hello1 reduce1hello1 reduce1spack1 reduce1spack1 reduce1

hadoop1 reduce1hello1 reduce1spack1 reduce1

spack1hadoop1hello1word1

进入Map端shuffle

hadoop1 reduce1hello1 reduce1spack1 reduce1spack1 reduce2word1 reduce2

spack1hadoop1hello1word1spack1

每个reduce去每个map里拿到自己分区的数据

合并排序

Map端shuffle结束

环形缓冲区