What's New in RapidMiner Radoop 8.0.0?
Released: December 04th, 2017
The following improvements are part of RapidMiner Radoop 8.0.0.
- New K-Means operator is available that is based on the Spark MLlib/ML clustering algorithm
- Hive access in Spark Script is now supported (HiveContext in Spark 1.x and SparkSession in Spark 2.x), if the user has the appropriate privileges on the cluster
- Added Spark 2.2 support (new option in the Spark Version list)
- Aggregate now support concatenation option as well
- Hive (JDBC) connection error now displays more information
- Import from Cluster Manager is now the default option to create a Radoop connection
- K-Means, Fuzzy K-Means and Canopy Mahout clustering operators are now deprecated
- Mahout test is now deprecated
- Replace missing quick fix is now offered if there are known missing values in the input of an operator that does not handle them
- BUGFIX: Installation of an Apache Spark release is no longer needed for more recent Cloudera Hadoop releases in order for Spark jobs to work (and not give an Unknown/unsupported param error). Use the new Spark 1.6 (CDH) Spark version connection setting in this case.
- BUGFIX: K-Means model apply inside Radoop Nest no longer fails with ClassNotFoundException or NoClassDefFoundError
- BUGFIX: Fixed Cannot initialize Cluster error when a certain failover proxy provider class is used on the cluster (org.apache.hadoop.yarn.client.RequestHedgingRMFailoverProxyProvider) and the connection is imported
- BUGFIX: Spark Script with R no longer relies on the private RDD API in Spark 2.x
- BUGFIX: Quick fix that adds a Type Conversion operator now selects the proper target type