Research on Parallel Support Vector Machine Based on Spark Big Data Platform

Table 2

Configuration of test cluster.

Number	Node name	Hadoop’s configuration	Spark’s configuration
1	Master	dfs.replication = 3; map.tasks.maximum = 16; reduce.tasks.maximum = 2; child.java.opts = -Xmx4096 M	SPARK_MEM = 20g
2	Slave01
3	Slave02
4	Slave03
5	Slave04
6	Slave05