If you choose to partition transactions according to user name, transactions entered by the same primary database user ID are processed serially. Only transactions entered by different user IDs are processed in parallel.
Use of this partitioning rule avoids contentions, but may in some cases cause unnecessary loss of parallelism. For example, consider a DBA who is running multiple batch jobs. If the DBA submits each batch job using the same user ID, Replication Server processes each one serially.
The user name partitioning rule is most useful if each user connection at the primary has a unique ID. It is less useful if multiple users log on using the same ID, such as “sa.” In such cases, orig_sessid may be a better option.