IBM FileNet P8, Version 5.2.1            

Tuning replication worker settings to improve performance

You can improve Content Federation Services for Image Services performance by adjusting the settings for inbound and outbound replication workers. The configuration settings apply to processing of both documents and annotations.

The default values are sufficient for most content federation configurations. You might need to modify the default settings when you synchronize a high volume of IBM® FileNet® Image Services documents, annotations, or a combination of both.

For more information about replication workers and instructions for how to access the configuration settings, see Configuring IBM FileNet Image Services replication workers.

There is no fixed formula to optimize Content Federation Services for Image Services performance tuning; each size must be uniquely tuned, based on the load and mix of operations, by using the following general guidelines. Setting each parameter to its highest performance setting does not always result in the best throughput. For optimal performance, try various combinations of the configuration parameters.

For more information about replication workers and transaction processing, see How replication workers process actions. Consider the following factors:

Adjusting the number of inbound and outbound replication workers

Increasing the number of replication workers increases the overall throughput of the import. However, increasing the number of replication workers has the following effects:
  • Increasing the number of replication workers increases the number of connections to the IBM FileNet Image Services server. Each replication worker has one connection to the IBM FileNet Image Services server. Monitor the number of IBM FileNet Image Services stubs and, if needed, raise the limit on the number of stubs.
  • Increasing the number of replication workers increases the Java™ virtual machine (JVM) memory that is used on the application server. You can monitor the application server heap size usage and increase the maximum heap size.
  • Carefully monitor the system vitals such as CPU, memory, and throughput rates. As system resources are maxed out, increasing the number of workers does not increase performance and might decrease performance.

Adjusting the number of documents per batch

The inbound batch size specifies the maximum number of documents that are replicated in a batch. Increasing this parameter causes each replication worker to replicate a larger number of IBM FileNet Image Services documents at any one time, requiring fewer round trips between the IBM FileNet Image Services and Content Platform Engine servers. However, making the batch size too large increases the risk that a replication worker does not have enough time to replicate an entire batch within the allotted IBM FileNet Image Services document services timeout period. If a batch is not fully replicated within the timeout period, it can be reprocessed by another IBM FileNet Image Services replication worker. A second IBM FileNet Image Services replication worker that reprocesses a batch that is not fully replicated results in duplication errors that are automatically rolled back. It also needlessly increases the CPU and database contention, due to the duplicate commit actions and subsequent rollback actions.

The outbound batch size specifies the maximum number of replication journals that are assigned to the outbound worker. Increasing this parameter can result in fewer round trips between the IBM FileNet Image Services and Content Platform Engine servers. Replication from Content Platform Engine to IBM FileNet Image Services entails copying the object content in addition to the replication of metadata; as a result, the outbound replication process takes longer to complete. You might be able to increase the outbound processing throughput with a larger outbound batch size. However, the CPU utilization also increases and a timeout might occur.

If you are running inbound and outbound batches simultaneously, configure the outbound batch size to be larger than the inbound batch size if you are using the default maximum number of inbound and outbound workers.

Adjusting the inbound and outbound batch delay period

The batch delay parameter specifies the number of seconds that the replication worker waits between batch requests. There are two batch delay parameters for each worker:
  • The busy wait interval is used if the worker just finished processing a batch. The default value for the busy wait interval is 0 seconds.
  • The idle wait interval is used if the worker just requested a batch and there is no data to process. The default value for the idle wait interval is 30 seconds.


Last updated: October 2015
p8ppt020.htm

© Copyright IBM Corporation 2015.