Hive使用双重GroupBy解决数据倾斜问题

1.数据准备

create table wordcount(a string) row format delimited fields terminated by ‘,’;

load data local inpath ‘opt/2.txt’ into table wordcount;

hive (default)> select * from wordcount;
OK
wordcount.a
b
a
a
a
a
b
b服务器托管网
c
c
e
d

2.双重group by实现 解决数据倾斜

随机数：ceil(rand()*10)

select split(sal服务器托管网t_a,‘‘)[1] alpah ,sum(count) from
(
select concat_ws(’’,cast(ceil(rand()*10) as string),a) salt_a,count(1) count from wordcount group by concat_ws(‘‘,cast(ceil(rand()*10) as string),a)
) b group by split(salt_a,’’)[1];

alpah _c1
a 4
b 3
c 2
d 1
e 1

解析:

2.1 第一层加盐group by

select concat_ws(‘‘,cast(ceil(rand()*10) as string),a) salt_a,count(1) count from wordcount group by concat_ws(’’,cast(ceil(rand()*10) as string),a)
salt_a count
10_a 1
10_b 1
1_a 2
2_a 1
3_b 1
4_b 1
4_c 1
4_d 1
6_c 1
7_e 1
Time taken: 176.729 seconds, Fetched: 10 row(s)

2.2 第二层去盐group by

select split(salt_a,‘_’)[1] alpah ,sum(count) from

b group by split(salt_a,‘_’)[1];

服务器托管，北京服务器托管，服务器租用 http://www.fwqtg.net

相关推荐: 排序嘉年华———归并排序

文章目录 一.归并是什么？ 题目一：合并有序数组 题目二：合并有序链表 二.归并排序 1.递归式归并 2.非递归式的归并排序 一.归并是什么？ 相信朋友们应该做过一类题，合服务器托管网并两个有序数组，在链表里也有合并两个单链表的oj题，那我们稍微回顾一下 题目…