hive中rcfile orcfile和parquetfile对比

来源:互联网 发布:面试技巧 知乎 编辑:程序博客网 时间:2024/06/07 09:39

一.开始创建三种格式的表:

create table rcfile (name string,age int,addr string,desc string) row format delimited fields terminated by ',' stored as rcfile;create table rcfile (name string,age int,addr string,desc string) row format delimited fields terminated by ',' stored as orcfile;create table rcfile (name string,age int,addr string,desc string) row format delimited fields terminated by ',' stored as parquetfile;

二.用shell生成1000W条数据,以”,”隔开,并且load data overwrite到 textfile表里面

三.分别把数据insert到三个表中:

insert into rcfile select * from lijie.textfile;insert into orcfile select * from lijie.textfile;insert into parquetfile select * from lijie.textfile;

四.开始测试

1.select * from xxfile           rcfile       Time taken: 47.604 seconds, Fetcheds 13756317 row(s)    orcfile      Time taken: 2.563 seconds, Fetcheds 13756317 row(s)    parquetfile  Time taken: 43.454 seconds, Fetcheds 13756317 row(s)结论orcfile 小于 rcfile 小于 parquet2.select name,addr from xxfile    rcfile       Time taken: 36.937 seconds, Fetcheds 13756317 row(s)    orcfile      Time taken: 2.514 seconds, Fetcheds 13756317 row(s)    parquetfile  Time taken: 43.454 seconds, Fetcheds 13756317 row(s)结论orcfile 小于 rcfile 小于 parquet3.select max(name) from xxfile    rcfile       Time taken: 34.375 seconds, Fetcheds 13756317 row(s)    orcfile      Time taken: 30.073 seconds, Fetcheds 13756317 row(s)    parquetfile  Time taken: 38.352 seconds, Fetcheds 13756317 row(s)结论orcfile 小于 rcfile 小于 parquet4.select count(1) from xxfile    rcfile       Time taken: 32.261 seconds, Fetcheds 13756317 row(s)    orcfile      Time taken: 28.959 seconds, Fetcheds 13756317 row(s)    parquetfile  Time taken: 32.265 seconds, Fetcheds 13756317 row(s)结论orcfile 小于 rcfile=parquet

五.总结
总数据量13756317
列:name,age,addr,desc
orcfile 查询效果更优,rcfile效果略好于parquetfile

0 0