Hive实践（hive0.12）

版本号：cdh5.0.0+hadoop2.3.0+hive0.12

一、原始数据：

1. 本地数据

[root@node33 data]# ll
total 12936
-rw-r--r--. 1 root root 13245467 May  1 17:08 hbase-data.csv
[root@node33 data]# head -n 3 hbase-data.csv 
1,1.52101,13.64,4.49,1.1,71.78,0.06,8.75,0,0,1
2,1.51761,13.89,3.6,1.36,72.73,0.48,7.83,0,0,1
3,1.51618,13.53,3.55,1.54,72.99,0.39,7.78,0,0,1

2. hdfs数据：

[root@node33 data]# hadoop fs -ls /input
Found 1 items
-rwxrwxrwx   1 hdfs supergroup   13245467 2014-05-01 17:09 /input/hbase-data.csv
[root@node33 data]# hadoop fs -cat /input/* | head -n 3
1,1.52101,13.64,4.49,1.1,71.78,0.06,8.75,0,0,1
2,1.51761,13.89,3.6,1.36,72.73,0.48,7.83,0,0,1
3,1.51618,13.53,3.55,1.54,72.99,0.39,7.78,0,0,1

二、创建hive表：

1.hive外部表：

[root@node33 hive]# cat employees_ext.sql 
create external table if not exists employees_ext(
	id	int,
	x1	float,
	x2	float,
	x3	float,
	x4	float,
	x5	float,
	x6	float,
	x7	float,
	x8	float,
	x9	float,
	y	int)
row format delimited fields terminated by ','
location '/input/'

创建表，client执行 ：hive -f employees_ext.sql

2. hive表

[root@node33 hive]# cat employees.sql 
create table employees(
	id	int,
	x1	float,
	x2	float,
	x3	float,
	x4	float,
	x5	float,
	x6	float,
	x7	float,
	x8	float,
	x9	float
)
partitioned by (y int);

创建表，client执行：hive -f employees.sql

3. hive表（orc方式存储）

[root@node33 hive]# cat employees_orc.sql 
create table employees_orc(
	id	int,
	x1	float,
	x2	float,
	x3	float,
	x4	float,
	x5	float,
	x6	float,
	x7	float,
	x8	float,
	x9	float
)
partitioned by (y int)
row format serde "org.apache.hadoop.hive.ql.io.orc.OrcSerde"
stored as orc;

执行：hive -f employees_orc.sql

三、导入数据：

1. employees_ext 表导入employees表：

[root@node33 hive]# cat employees_ext-to-employees.sql 

set hive.exec.dynamic.partition=true;
set hive.exec.dynamic.partition.mode=nonstrict;
set hive.eec.max.dynamic.partitions.pernode=1000;

insert overwrite table employees 
	partition(y)
select 
	emp_ext.id,
	emp_ext.x1,
	emp_ext.x2,
	emp_ext.x3,
	emp_ext.x4,
	emp_ext.x5,
	emp_ext.x6,
	emp_ext.x7,
	emp_ext.x8,
	emp_ext.x9,
	emp_ext.y
from employees_ext emp_ext;

执行：hive -f employees_ext-to-employees.sql。其部分log例如以下：

Partition default.employees{y=1} stats: [num_files: 1, num_rows: 0, total_size: 3622, raw_data_size: 0]
Partition default.employees{y=2} stats: [num_files: 1, num_rows: 0, total_size: 4060, raw_data_size: 0]
Partition default.employees{y=3} stats: [num_files: 1, num_rows: 0, total_size: 910, raw_data_size: 0]
Partition default.employees{y=5} stats: [num_files: 1, num_rows: 0, total_size: 699, raw_data_size: 0]
Partition default.employees{y=6} stats: [num_files: 1, num_rows: 0, total_size: 473, raw_data_size: 0]
Partition default.employees{y=7} stats: [num_files: 1, num_rows: 0, total_size: 13561851, raw_data_size: 0]
Table default.employees stats: [num_partitions: 6, num_files: 6, num_rows: 0, total_size: 13571615, raw_data_size: 0]
MapReduce Jobs Launched: 
Job 0: Map: 1   Cumulative CPU: 6.78 sec   HDFS Read: 13245660 HDFS Write: 13571615 SUCCESS
Total MapReduce CPU Time Spent: 6 seconds 780 msec

OK
Time taken: 186.743 seconds

查看hdfs文件大小：

[root@node33 hive]# hadoop fs -count /user/hive/warehouse/employees
           7            6           13571615 /user/hive/warehouse/employees

查看hdfs文件内容：

bash-4.1$ hadoop fs -cat /user/hive/warehouse/employees/y=1/* | head -n 1
11.5210113.644.491.171.780.068.750.00.0

（截图的内容为输出，拷贝到代码块里面有问题）

2. employees_ext 表导入employees_orc表：

[root@node33 hive]# cat employees_ext-to-employees_orc.sql 

set hive.exec.dynamic.partition=true;
set hive.exec.dynamic.partition.mode=nonstrict;
set hive.eec.max.dynamic.partitions.pernode=1000;

insert overwrite table employees_orc 
	partition(y)
select 
	emp_ext.id,
	emp_ext.x1,
	emp_ext.x2,
	emp_ext.x3,
	emp_ext.x4,
	emp_ext.x5,
	emp_ext.x6,
	emp_ext.x7,
	emp_ext.x8,
	emp_ext.x9,
	emp_ext.y
from employees_ext emp_ext;

执行：hive -f employees_ext-to-employees_orc.sql，其部分log例如以下：

Partition default.employees_orc{y=1} stats: [num_files: 1, num_rows: 0, total_size: 2355, raw_data_size: 0]
Partition default.employees_orc{y=2} stats: [num_files: 1, num_rows: 0, total_size: 2539, raw_data_size: 0]
Partition default.employees_orc{y=3} stats: [num_files: 1, num_rows: 0, total_size: 1290, raw_data_size: 0]
Partition default.employees_orc{y=5} stats: [num_files: 1, num_rows: 0, total_size: 1165, raw_data_size: 0]
Partition default.employees_orc{y=6} stats: [num_files: 1, num_rows: 0, total_size: 955, raw_data_size: 0]
Partition default.employees_orc{y=7} stats: [num_files: 1, num_rows: 0, total_size: 1424599, raw_data_size: 0]
Table default.employees_orc stats: [num_partitions: 6, num_files: 6, num_rows: 0, total_size: 1432903, raw_data_size: 0]
MapReduce Jobs Launched: 
Job 0: Map: 1   Cumulative CPU: 7.84 sec   HDFS Read: 13245660 HDFS Write: 1432903 SUCCESS
Total MapReduce CPU Time Spent: 7 seconds 840 msec
OK
Time taken: 53.014 seconds

查看hdfs文件大小：

[root@node33 hive]# hadoop fs -count /user/hive/warehouse/employees_orc
           7            6            1432903 /user/hive/warehouse/employees_orc

查看hdfs文件内容：

 

3. 比較两者性能

 

	时间	压缩率
employees表：	186.7秒	13571615/13245660=1.0246
employees_orc表：	53.0秒	1432903/13245660=0.108

时间上来说，orc的表现方式会好非常多。同一时候压缩率也好非常多。

只是，这个測试是在本人虚拟机上測试的，并且是单机測试的，所以參考价值不是非常大，可是压缩率还是有一定參考价值的。

四、导出数据

1. employees表：

[root@node33 hive]# cat export_employees.sql 

insert overwrite local directory '/opt/hivedata/employees.dat'
row format delimited
fields terminated by ','
select 
	emp.id,
	emp.x1, 
	emp.x2, 
	emp.x3, 
	emp.x4, 
	emp.x5, 
	emp.x6, 
	emp.x7, 
	emp.x8, 
	emp.x9, 
	emp.y
from employees emp

执行：hive -f export_employees.sql
部分log：

MapReduce Total cumulative CPU time: 9 seconds 630 msec
Ended Job = job_1398958404577_0007
Copying data to local directory /opt/hivedata/employees.dat
Copying data to local directory /opt/hivedata/employees.dat
MapReduce Jobs Launched: 
Job 0: Map: 1   Cumulative CPU: 9.63 sec   HDFS Read: 13572220 HDFS Write: 13978615 SUCCESS
Total MapReduce CPU Time Spent: 9 seconds 630 msec
OK
Time taken: 183.841 seconds

数据查看：

[root@node33 hive]# ll /opt/hivedata/employees.dat/
total 13652
-rw-r--r--. 1 root root 13978615 May  2 05:15 000000_0
[root@node33 hive]# head -n 1 /opt/hivedata/employees.dat/000000_0 
1,1.52101,13.64,4.49,1.1,71.78,0.06,8.75,0.0,0.0,1

2. employees_orc表：

[root@node33 hive]# cat export_employees_orc.sql 

insert overwrite local directory '/opt/hivedata/employees_orc.dat'
row format delimited
fields terminated by ','
select 
	emp.id,
	emp.x1, 
	emp.x2, 
	emp.x3, 
	emp.x4, 
	emp.x5, 
	emp.x6, 
	emp.x7, 
	emp.x8, 
	emp.x9, 
	emp.y
from employees_orc emp

执行 hive -f export_employees_orc.sql

部分log：

MapReduce Total cumulative CPU time: 4 seconds 920 msec
Ended Job = job_1398958404577_0008
Copying data to local directory /opt/hivedata/employees_orc.dat
Copying data to local directory /opt/hivedata/employees_orc.dat
MapReduce Jobs Launched: 
Job 0: Map: 1   Cumulative CPU: 4.92 sec   HDFS Read: 1451352 HDFS Write: 13978615 SUCCESS
Total MapReduce CPU Time Spent: 4 seconds 920 msec
OK
Time taken: 41.686 second

查看数据：

[root@node33 hive]# head -n 1 /opt/hivedata/employees_orc.dat/000000_0 
1,1.52101,13.64,4.49,1.1,71.78,0.06,8.75,0.0,0.0,1
[root@node33 hive]# ll /opt/hivedata/employees_orc.dat/
total 13652
-rw-r--r--. 1 root root 13978615 May  2 05:18 000000_0

这里的数据和原始数据的大小不一样。原始数据是13245467， 而导出到本地的是13978615 。这是由于数据的精度问题，比如原始数据中的0都被存储为了0.0。

 

分享，成长。快乐

转载请注明blog地址：http://blog.csdn.net/fansy1990