How to configure a J2EE cluster for PI 7.3* system to achieve better performance in the Java components?
Listed below are the two general approaches for the configuration of more J2EE server processes:
1) Additional Server Nodes On Central Instance
The user can add additional server nodes to the central instance, Incase, your host has plenty of resources. This is the favoured configuration, as the costs of configuration and administration are comparatively lower in this case.
- Each J2EE server node assigns the quantified MaxHeapSize, and the memory consumption is totalled across all server nodes as a result. User does not set the heap size of the individual J2EE server nodes too high in this aspect.
- If the user utilizes many J2EE server nodes, the user can reduce the number of parallel JAVA mapping connections (JCo RFC provider - destination AI_RUNTIME_<SID>) for each server node.
The JCo RFC provider destination exists only in the Adapter Engine of a dual-stack PI system, however not in a Java-only PI system (AEX) as the PI processing is locally performed in the Adapter Engine in this scenario.
- If the user has already made local changes to the parameterization in his existing J2EE server node, user should ensure that these parameter changes are also transferred to the new J2EE server node. The easiest way of doing this is by always performing the parameterization globally (Global Configuration).
- A new instance of the defined PI communication channels is created, with each server node. This response should be considered when the user defines the channels and program the connection or when the user programs the adapter modules.
For instance :
- EOIO (Exactly Once in Order) channels are active only on one server node; additionally, the user can specify the cluster ID of a J2EE server node in the channel configuration.
- EO (Exactly Once) channels which connect with JMS queues are active on all nodes and process messages in parallel.
- Regardless of the QoS configuration Channels which connect with JMS topics are active on one node only.
- Point 4: FTP Sender File Processing in Cluster Environment
- Point 7: File Construction Mode "Add Time Stamp" and two general approaches for the configuration of more J2EE server processes in Cluster Environment
- Point 44: File Receiver with append mode on multiple cluster nodes
- Point 48: File System(NFS) File Sender Processing in Cluster Environment
User must ensure that read entries in the database tables are either deleted immediately or are marked as being transferred.
2) Additional Dialog Instances
The user can install an additional dialog instance and one or more server nodes.
- Each J2EE server node assigns the specified MaxHeapSize and the memory consumption is thereby totalled across all servers on one host. Therefore, user should not set the heap size as very high.
- The above points for AFW, mapping, and adapters are also applicable to this installation.