SQL*Net round trip tracing is a useful tool for helping determine or eliminate possible sources of network latency.  As the name implies, a round trip is the number of trips required to satisfy a request for data.  An excessive number of round trips could cause network performance issues.   Many factors can impact the number of round trips.

For example, ARRAYSIZE can impact network performance.  The ARRAYSIZE determines how much data can be returned in each round trip; therefore, an ARRAYSIZE that is too small for the data requested can force too many round trips.  To optimized the ratio of round trips, the ARRAYSIZE can be adjusted.

Network buffers and packets can also impact the number of round trips.  While it's still true that tuning of network packet traffic happens at the network level, it is also important that the Oracle net parameter be optimized as well.

The session data unit (SDU) and transport data unit (TDU) are resident in the tnsnames.ora file and can have an impact on round trips.  The SDU determines the packet size to be sent over the network while the TDU is the default packet size for grouping data together.  There are some important guidelines for setting the SDU and TDU properly in the tnsnames.ora.  Also, see these notes for Optimizing Oracle Net parameters.

The SQL*Net more data to client  wait event happens when Oracle writes multiple data buffers (sized per SDU) in a single logical network call; however, it is not a direct measure of network latency.

In order to determine the appropriate settings, it will be helpful to trace the SQL*Net round trips as empirical tests are conducted.  Tips on how to initiate SQL*Net traces from both the client and server side are available HERE.

本文转自海天一鸥博客园博客，原文链接http://www.cnblogs.com/sgsoft/archive/2010/12/15/1907356.html，如需转载请自行联系原作者