Replication

You can use the replication feature in a directory server to improve performance, availability, and reliability of the server. The replication process synchronizes data that are stored in multiple directory servers.

By using multi-threaded replication (asynchronous), you can activate replication to use multiple threads during the replication process. Multi-threaded replication improves the overall throughput of replication.

Multi-threaded replication is difficult to administer if servers or networks are not reliable. You can consider switching to multi-threaded replication for the following replication environments:

• A high update rate
• No previous versions of directory servers
• Common Advanced Encryption Standard (AES) salt and synchronization if the encryption is AES and passwords are updated often
• Small fanout (for example, eight connections per agreement with 24 replicas might be too complicated depending on system configuration)
• Available servers and reliable network
• Real-time data consistency is not critical
• All replication schedules are immediate
• Multiprocessor systems

If replication errors occur, the errors are logged in the log files. You must monitor the error logs and fix any replication issues. To search for the replication backlog for all agreements that are supplied by a server, run the following command:

ldapsearch -h sup_host -p port -D adminDN -w ? -s sub -b repl_context\
	objectclass=ibm-replicationagreement 	ibm-replicationpendingchangecount ibm-replicationstate

If the replication state is active and the pending count is growing, then the replication backlog might not decrease. For the replication backlog to decrease, the update rate must decrease or change the replication mode from synchronous to asynchronous (multi-threaded).

Replication also adds to the workload on the master server where the updates are first applied. In addition to updating its copy of the directory data, the master server sends the changes to all replica servers. If your application or users do not depend on immediate replication, then you must avoid setting the replication schedule at peak activity times. Scheduling replication at idle time can minimize the degrading the throughput on the master server.

You can tune the following configurable values to improve the replication performance:

• Number of replication threads per supplier and consumer
• Replication context cache size
• Replication ready size limit