Asynchronous data refreshes can consume large amounts of memory and inflict disk I/O overhead. To avoid this, you can use a synchronous, messageless Advanced Replication interface. In this mode, the associated data is sent as a data stream, thus conserving memory. This mode also reduces the needed disk I/O bandwidth, because no messages are written to disk.

You can define how the replica table is being locked in conjunction with the asynchronous data refresh, see Pessimistic and optimistic concurrency controls.

By default, all data in a publication refresh is written to the replica database in a single transaction. If the reply to the refresh request contains refreshes of large publications, the size of the commit block, that is, the number of rows that are committed in one transaction, can be defined using the COMMITBLOCK property.

Setting the maximum size of the commitblock can improve performance of the replica databases. However, data integrity cannot be guaranteed if the data is transmitted in more than one transaction and active users on the replica are changing data at the same time. Therefore, it is recommended that you disconnect all online users from the replica database when you use the COMMITBLOCK option.

The TIMEOUT property defines how long the replica server will wait for the reply message. If TIMEOUT is not defined, then FOREVER is used.

The following set of statements provide an example of a synchronous refresh:

For more information about the synchronous refresh syntax, see REFRESH.