Choosing Packet Sizes

For best performance, choose a server packet size that works efficiently with the underlying packet size on your network.

The goals are:

Reducing the number of server reads and writes to the network
Reducing unused space in network packets (increasing network throughput)

For example, if your network packet size carries 1500 bytes of data, setting the SAP ASE packet size to 1024 (512*2) will probably achieve better performance than setting it to 1536 (512*3). The figure below shows how four different packet size configurations would perform in such a scenario.

Factors in determining packet size

Graphic showing four bar graphs describing the factors used in determining the best packet size. Bar graph one, with a packet size of 512, has an acceptable level of performance with 68% of the packet used. Bar graph 2 with a packet size of 1024 has a similar performance. Bar graph 3, with a packet size of 2560, has the best performance, with 85% of the packet used. Bar graph 4, with a packet size of 1536 has the worst performance, with 51% used. All packets have the same overhead: 1500 bytes.

After you determine the available data space of the underlying packets on your network, perform your own benchmark tests to determine the optimum size for your configuration.

Use sp_sysmon to see how changing max network packet size affects network I/O management and task switching. For example, try increasing max network packet size and then checking sp_sysmon output to see how this affects bcp for large batches. See the Performance and Tuning Series: Monitoring SAP Adaptive Server with sp_sysmon.