Index: docs/xdocs/user_manual/cassandra_storage_handler.xml
===================================================================
--- docs/xdocs/user_manual/cassandra_storage_handler.xml (revision 0)
+++ docs/xdocs/user_manual/cassandra_storage_handler.xml (revision 0)
@@ -0,0 +1,84 @@
+
+
+
+ Jar files found in the auxlib directory are automatically added to the classpath
+of map and reduce tasks. Creating an auxlib folder under $hive_home automatically
+loads these classes. Place all the jars from your cassandra distribution (except the hadoop jar)
+into auxlib. In order to interact with Cassandra you must define a table within hive. This table
+must be an EXTERNAL TABLE. The Cassandra Hive integration requires your storage-xml to be on the map/reduce tasks classpath.
+One way to accomplish this is by using the add file capability in hive to add the file to the classpath. Apache Cassandra is an NoSQL store.The Hive Cassandra integration was built against Apache Cassandra 6.3. Currently no super-column support exists. The client to Cassandra is Thrift generated Java bindings. The Cassandra Storage Handler uses these for underlying communication to Cassandra. Cassandra provides hadoop support with the ColumnFamilyInputFormat class. A sample mapper for Cassandra ConfigurationHelper helps set configuration variables such as:
+Create an auxlib
+Example of Writing Data
+
+
+
The ColumnFamilyInputFormat is built against the mapreduce.* api not the mapred.* api. HiveCassandraTableInputFormat is an +overlay on ColumnFamilyInputFormat to address this.
+ +The types of the ColumnFamilyInputFormat String, SortedMap byte[], and IColumn value are not serializable. In most +cases hive ignores the key from InputFormats and only uses the value. HiveIColumn implements the Cassandra IColumn and Writable +interfaces. CassandraRowResult stores the key and value into a single object.
+ +As of Apache Cassandra (6.3), Cassandra currently does not have an OutputFormat class. Additionally insertion is done via +the method call described above. There is no Writable object to abstract an insert. This is +solved using the CassandarPut class. CassandraPut encapsulates row key, the columns to be added +for that row key are Serialized instances of the CassandraColumn class. The HiveCassandraOutputFormat +is fed Text, CassandraPut pairs through Map Reduce and writes them via the thrift interface.
+The Cassandra SerDe currently allows ONLY external tables to be specified. This is because the +6.X does not allow meta operations. The Cassandra SedDe requires parameters to set to work. These +include:
+