Beginning with CDH 5.7, Hive-on-Spark is supported, offering native support for Spark as the execution engine for Hive in addition to MapReduce. Hive-on-Spark provides better performance than Hive-on-MapReduce for most use cases, with no loss of functionality. Switching to Hive-on-Spark requires no changes on user queries, including UDFs and other semantics.
Starting with CDH 5.7, partners certifying with Hive will be required to test with Hive-on-Spark. Partners who are already certified with Hive-on-MapReduce prior to CDH5.7 are invited and encouraged to recertify on CDH 5.7, in order to provide customer reassurance that the product has been tested with Hive-on-Spark.
For tips on making Hive-on-Spark work well, see the CDH 5.7 draft documentation (specifically the section entitled, “Tuning Hive-on-Spark”) on the Cloudera Connect Partner Portal, .