Join Optimization Join Optimization Improvements to the Hive Optimizer Star Join Optimization Star Schema Example Prior Support for MAPJOIN Limitations of Prior Implementation Enhancements for Star Joins Optimize Chains of Map Joins Current and Futur
hive的多表连接,都会转换成多个MR job,每一个MR job在hive中均称为Join阶段.按照join程序最后一个表应该尽量是大表,因为join前一阶段生成的数据会存在于Reducer 的buffer中,通过stream最后面的表,直接从Reducer中读取已经缓冲的中间数据结果,与后面的大表进行连接时,只需要从buffer中读取缓存的key,与大表中的指定key进行连接,速度更快,也避免内存缓冲区溢出. SELECT a.val, b.val, c.val FROM a JOIN b