hive中的bucket table

前言

bucket table(桶表)是对数据进行哈希取值，然后放到不同文件中存储

应用场景

当数据量比较大，我们需要更快的完成任务，多个map和reduce进程是唯一的选择。
但是如果输入文件是一个的话，map任务只能启动一个。
此时bucket table是个很好的选择，通过指定CLUSTERED的字段，将文件通过hash打散成多个小文件。

create table test
(id int,
 name string
)
CLUSTERED BY(id) SORTED BY(name) INTO 32 BUCKETS
ROW FORMAT DELIMITED
FIELDS TERMINATED BY ‘/t’;

执行insert前不要忘记设置

set hive.enforce.bucketing = true;

强制采用多个reduce进行输出

hive> INSERT OVERWRITE TABLE test select * from test09;
Total MapReduce jobs = 1
Launching Job 1 out of 1
Number of reduce tasks determined at compile time: 32
In order to change the average load for a reducer (in bytes):
  set hive.exec.reducers.bytes.per.reducer=<number>
In order to limit the maximum number of reducers:
  set hive.exec.reducers.max=<number>
In order to set a constant number of reducers:
  set mapred.reduce.tasks=<number>
Starting Job = job_201103070826_0018, Tracking URL = http://hadoop00:50030/jobdetails.jsp?jobid=job_201103070826_0018
Kill Command = /home/hjl/hadoop/bin/../bin/hadoop job  -Dmapred.job.tracker=hadoop00:9001 -kill job_201103070826_0018
2011-03-08 11:34:23,055 Stage-1 map = 0%,  reduce = 0%
2011-03-08 11:34:27,084 Stage-1 map = 6%,  reduce = 0%
*************************************************
Ended Job = job_201103070826_0018
Loading data to table test
5 Rows loaded to test
OK
Time taken: 175.036 seconds

hive的sunwg_test11文件夹下面出现了32个文件，而不是一个文件

[hadoop@hadoop00 ~]$ hadoop fs -ls /ticketdev/test
Found  items
-rw-r–r–    ticketdev hadoop           -- : /ticketdev/test/attempt_201103070826_0018_r_000000_0
-rw-r–r–    ticketdev hadoop           -- : /ticketdev/test/attempt_201103070826_0018_r_000001_0
-rw-r–r–    ticketdev hadoop           -- : /ticketdev/test/attempt_201103070826_0018_r_000002_0
-rw-r–r–    ticketdev hadoop           -- : /ticketdev/test/attempt_201103070826_0018_r_000003_0
-rw-r–r–    ticketdev hadoop           -- : /ticketdev/test/attempt_201103070826_0018_r_000004_0
-rw-r–r–    ticketdev hadoop           -- : /ticketdev/test/attempt_201103070826_0018_r_000005_0
-rw-r–r–    ticketdev hadoop           -- : /ticketdev/test/attempt_201103070826_0018_r_000006_0
-rw-r–r–    ticketdev hadoop           -- : /ticketdev/test/attempt_201103070826_0018_r_000007_0
-rw-r–r–    ticketdev hadoop           -- : /ticketdev/test/attempt_201103070826_0018_r_000008_0
-rw-r–r–    ticketdev hadoop           -- : /ticketdev/test/attempt_201103070826_0018_r_000009_0
-rw-r–r–    ticketdev hadoop           -- : /ticketdev/test/attempt_201103070826_0018_r_000010_0
-rw-r–r–    ticketdev hadoop           -- : /ticketdev/test/attempt_201103070826_0018_r_000011_0
-rw-r–r–    ticketdev hadoop           -- : /ticketdev/test/attempt_201103070826_0018_r_000012_0
-rw-r–r–    ticketdev hadoop           -- : /ticketdev/test/attempt_201103070826_0018_r_000013_0
-rw-r–r–    ticketdev hadoop           -- : /ticketdev/test/attempt_201103070826_0018_r_000014_0
-rw-r–r–    ticketdev hadoop           -- : /ticketdev/test/attempt_201103070826_0018_r_000015_0
-rw-r–r–    ticketdev hadoop           -- : /ticketdev/test/attempt_201103070826_0018_r_000016_0
-rw-r–r–    ticketdev hadoop           -- : /ticketdev/test/attempt_201103070826_0018_r_000017_0
-rw-r–r–    ticketdev hadoop           -- : /ticketdev/test/attempt_201103070826_0018_r_000018_0
-rw-r–r–    ticketdev hadoop           -- : /ticketdev/test/attempt_201103070826_0018_r_000019_0
-rw-r–r–    ticketdev hadoop           -- : /ticketdev/test/attempt_201103070826_0018_r_000020_0
-rw-r–r–    ticketdev hadoop           -- : /ticketdev/test/attempt_201103070826_0018_r_000021_0
-rw-r–r–    ticketdev hadoop           -- : /ticketdev/test/attempt_201103070826_0018_r_000022_0
-rw-r–r–    ticketdev hadoop           -- : /ticketdev/test/attempt_201103070826_0018_r_000023_0
-rw-r–r–    ticketdev hadoop           -- : /ticketdev/test/attempt_201103070826_0018_r_000024_0
-rw-r–r–    ticketdev hadoop           -- : /ticketdev/test/attempt_201103070826_0018_r_000025_0
-rw-r–r–    ticketdev hadoop           -- : /ticketdev/test/attempt_201103070826_0018_r_000026_0
-rw-r–r–    ticketdev hadoop           -- : /ticketdev/test/attempt_201103070826_0018_r_000027_0
-rw-r–r–    ticketdev hadoop           -- : /ticketdev/test/attempt_201103070826_0018_r_000028_0
-rw-r–r–    ticketdev hadoop           -- : /ticketdev/test/attempt_201103070826_0018_r_000029_0
-rw-r–r–    ticketdev hadoop           -- : /ticketdev/test/attempt_201103070826_0018_r_000030_0
-rw-r–r–    ticketdev hadoop           -- : /ticketdev/test/attempt_201103070826_0018_r_000031_0

文件被打散后，可以启动多个mapreduce task
当执行一些操作的时候，你会发现系统启动了32个map任务