Packet size determines how many bytes are sent in each Replication Server command buffer. See Table A-2 for packet size information.
Packet size |
Description |
|---|---|
Packet_size=512 |
Minimum size; typically only useful for test purposes. |
Packet_size=32K |
Packet sizes in increments of 1024 bytes are required when packet size is greater than 1024 bytes. |
Packet_size=32000 |
Use either whole numbers or abbreviate, using the K (1000) symbol. |
Packet_size=1024K |
Maximum packet size. |
Packet size is useful when tuning Replication Agent for efficient operation in a network environment. Packet size must be larger than the largest command sent to Replication Server. A DB2 row can be as large as 32K. Determining the best packet size depends on configurations that change the size of data in a command, such as minimal columns. Inserts and deletes have a maximum size of ~32K; however, the before and after columns of an update make the maximum data size ~64K.
A minimum size of 175 bytes is needed for a typical Replication Server command. You must also allow enough space for table names, column names, punctuation, and other information necessary for a Replication Server command. If packet size is exceeded by a single command, the log extract terminates with a user abend of 1408.
Configuration packet size should be tested in a production-like environment. When testing, use the same version of Replication Server and a substantive amount of data. Changing the setting should not affect Replication Agent functionality, but it is important to notice how the entire replication environment is affected. Increasing packet size sends data over the network in larger packets, with fewer sends and receives. Latency, the frequency with which truncation points are requested, and other configuration issues should also be considered when modifying packet size.
