##File Packing List The CCIndex V1.1.2 package consists of:

• ict.ocrabase.main.java.client
• ict.ocrabase.main.java.regionserver
• ict.ocrabase.main.java.test
• org.apache.hadoop.hbase.filter
• org.apache.hadoop.hbase.filter
• org.apache.hbase.coprocessor

The total overview like this:

##Installation Prerequisites: the Hadoop and the HBase environment have been right configured. the hadoop version is hadooop-2.5.1. the hbase version is hbase-1.1.2

• Copy the file CCIndex-1.1.2.jar to path : $HADOOP_ENV$/share/hadoop/yarn . like this:

• Add the src file to your project if you want to debug your code.(if not, you should add the CCIndex-1.1.2.jar to your path )
• Add the flowing configuration to $HBASE-ENV$/conf/hbase-site.xml

    <property>
      <name>hbase.coprocessor.region.classes</name>             
      <value>org.apache.hbase.coprocessor.PutObserver,org.apache.hbase.coprocessor.DeleteObserver</value>
    </property>

Up to now, the environment have been successfully configured

##How to use There are two ways to use CCIndex. One is no bulkload, the other is bulkload. ###No bulkload No bulkload means you have to put data to you table, when there are a large amount of data, it may consume huge time. There have some example in the ict.ocrabase.main.java.test package.

Example:

ict.ocrabase.main.java.test.CreateTableWithIndexTest.java create hbase table and the ccindex table. after accomplish, you will see there exists three tables in hbase like this: ict.ocrabase.main.java.test.PutTableWithIndexTest.java put data to the hbase table and the ccindex table ###Bulkload Bulkload use map reduce to put data to the base table and the ccindex table

Example:

• First upload the TPC-H test data to the hdfs.download

bin/hadoop fs -put ../test-data/xaa1000.txt /index-data

• ict.ocrabase.main.java.test.CreateTableTestBulkload.java create hbase table and the ccindex table. after run the above code, you will see there exists three tables like this:

• import data ict.ocrabase.main.java.client.cli.Import.java import the hdfs data to hbase, the configuration parameter like this:

-s /index-data -ts real_table_with_index,SEMICOLON,f:c1:STRING:CCINDEX:real_table_with_index-f_c1,f:c2:STRING,f:c3:STRING,f:c4:STRING:CCINDEX
:real_table_with_index-f_c4,f:c5:STRING,f:c6:STRING,f:c7:STRING,f:c8:STRING -l 32

after run the above code, you will see some output like this: PS: there the index columns are c1 and c4 ##Query

• Scan without index Example:

ict.ocrabase.main.java.test.QueryByCondition.java

• Scan with index Example:

ict.ocrabase.main.java.test.QueryMultiColumnUseCCIndex.java ##Simple query test

• Data source: TPC-H ORDERS table
• Sample data source in our test download
• Total 10 million line data
• Index column c1(CUSTKEY) and c4(ORDERDATE)
• The query sql like this:

select * from ORDERS where 100000<CUSTKEY<300000 and "1994-01-01"<ORDERDATE<"1994-01-30"

###caching number=1,threads number=1 tbale1

tbale2

Table1 only use c4 as query condition, the result count are 33514; Table2 use both c1 and c4 as query condition the result count are 13649; ###caching number=10000,threads number=10

tbale1

tbale2

CCIndex: A Complemental Clustering Index on Distributed Ordered Tables for Multi-dimensional Range Queries