IBM Open Platform – Recommended Services Layout for Hadoop


Non-HA Deployment

We recommend a minimum of one (1) Management node for Development and Test clusters (when performance is not a factor). If performance is a concern, we recommend a minimum of three (3) Management nodes. If Big SQL is used, we recommend a minimum of four (4) Management nodes.

Management Node 1
 Ambari (PostgreSQL)
 PostgreSQL
 Nagios
 Ganglia
 Knox
 Journal Node
 Zookeeper
 Hive

Management Node 2
 Resource Manager
 Hbase Master,
 Journal Node
 Zookeeper
 Oozie

Management Node 3
 Name Node
 Job history server
 Journal Node
 Zookeeper

Management Node 4
 Big sql Headnode
 Big sql Scheduler,
 Hive Server (MySQL)
 MySQL metastore

HA Deployment

We recommend six (6) Management nodes.

Management Node 1

 Ambari
 PostgreSQL
 Nagios
 Ganglia

Management Node 2
 Resource Manager
 Name Node (standby)
 Journal Node
 Zookeeper
 Oozie

Management Node 3
 Resource Manager (standby)
 Name Node
 Job history server
 Journal Node
 Zookeeper

Management Node 4
 Big sql Headnode
 Big sql Scheduler
 Hbase Master (standby)
 Hive Server
 MySQL metastore

Management Node 5
 Big sql Headnode (Standby)
 Big sql Scheduler (Standby)
 Hbase Master
 Hive Server
 Journal Node
 Zookeeper

Management Node 6
 Knox