Quote: Instead of storing intermediate data after a computation into on-disk storage such as HDFS,
these data can be stored "fault tolerantly" in-memory.
The ability to store intermediate data in-memory
across multiple nodes is one of the reasons
why Spark is able to execute faster than Hadoop's MapReduce.
BRON
mrt 2016
Laatst gewijzigd door Nr.10 : 27 mei 2017 om 02:14.
|