LZO Compression

General LZO Concepts

LZO is a lossless data compression library that favors speed over compression ratio. See http://www.oberhumer.com/opensource/lzo and http://www.lzop.org for general information about LZO and see Compressed Data Storage for information about compression in Hive.

Imagine a simple data file that has three columns

  • id
  • first name
  • last name

Let's populate a data file containing 4 records:

19630001     john          lennon
19630002     paul          mccartney
19630003     george        harrison
19630004     ringo         starr

Let's call the data file /path/to/dir/names.txt.

In order to make it into an LZO file, we can use the lzop utility and it will create a names.txt.lzo file.

Now copy the file names.txt.lzo to HDFS.

Prerequisites

Lzo/Lzop Installations

lzo and lzop need to be installed on every node in the Hadoop cluster. The details of these installations are beyond the scope of this document.

core-site.xml

Add the following to your core-site.xml:

  • com.hadoop.compression.lzo.LzoCodec
  • com.hadoop.compression.lzo.LzopCodec

For example:

<property>
<name>io.compression.codecs</name>
<value>org.apache.hadoop.io.compress.GzipCodec,org.apache.hadoop.io.compress.DefaultCodec,org.apache.hadoop.io.compress.BZip2Codec,com.hadoop.compression.lzo.LzoCodec,com.hadoop.compression.lzo.LzopCodec</value>
</property>

<property>
<name>io.compression.codec.lzo.class</name>
<value>com.hadoop.compression.lzo.LzoCodec</value>
</property>

Next we run the command to create an LZO index file:

hadoop jar /path/to/jar/hadoop-lzo-cdh4-0.4.15-gplextras.jar com.hadoop.compression.lzo.LzoIndexer  /path/to/HDFS/dir/containing/lzo/files

This creates names.txt.lzo on HDFS.

Table Definition

The following hive -e command creates an LZO-compressed external table:

hive -e "CREATE EXTERNAL TABLE IF NOT EXISTS hive_table_name (column_1  datatype_1......column_N datatype_N)
         PARTITIONED BY (partition_col_1 datatype_1 ....col_P  datatype_P)
         ROW FORMAT DELIMITED FIELDS TERMINATED BY '\t'
         STORED AS INPUTFORMAT  \"com.hadoop.mapred.DeprecatedLzoTextInputFormat\"
                   OUTPUTFORMAT \"org.apache.hadoop.hive.ql.io.HiveIgnoreKeyTextOutputFormat\";

Note: The double quotes have to be escaped so that the 'hive -e' command works correctly.

See CREATE TABLE and Hive CLI for information about command syntax.

Hive Queries

Option 1: Directly Create LZO Files

  1. Directly create LZO files as the output of the Hive query.
  2. Use lzop command utility or your custom Java to generate .lzo.index for the .lzo files.

Hive Query Parameters

SET mapreduce.output.fileoutputformat.compress.codec=com.hadoop.compression.lzo.LzoCodec
SET hive.exec.compress.output=true
SET mapreduce.output.fileoutputformat.compress=true

For example:

hive -e "SET mapreduce.output.fileoutputformat.compress.codec=com.hadoop.compression.lzo.LzoCodec; SET hive.exec.compress.output=true;SET mapreduce.output.fileoutputformat.compress=true; <query-string>"

     Note: If the data sets are large or number of output files are large , then this option does not work.

Option 2: Write Custom Java to Create LZO Files

  1. Create text files as the output of the Hive query.
  2. Write custom Java code to
    1. convert Hive query generated text files to .lzo files
    2. generate .lzo.index files for the .lzo files generated above

Hive Query Parameters

Prefix the query string with these parameters:

SET hive.exec.compress.output=false
SET mapreduce.output.fileoutputformat.compress=false

For example:

hive -e "SET hive.exec.compress.output=false;SET mapreduce.output.fileoutputformat.compress=false;<query-string>"
  • No labels