Using stored procedures, views, and triggers can reduce network traffic. These Transact-SQL tools can store large chunks of code on the server so that only short commands need to be sent across the network.
Stored procedures – applications that send large batches of Transact-SQL commands may place less load on the network if the SQL is converted to stored procedures. Views can also help reduce the amount of network traffic.
You may be able to reduce network overhead by turning off doneinproc packets.
Ask for only the information you need – applications should request only the rows and columns they need, filtering as much data as possible at the server to reduce the number of packets that need to be sent. In many cases, this can also reduce the disk I/O load.
Large transfers – simultaneously decrease overall throughput and increase the average response time. If possible, perform large transfers during off-hours. If large transfers are common, consider acquiring network hardware that is suitable for such transfers. Table 2-1 shows the characteristics of some network types.
Type |
Characteristics |
---|---|
Token ring |
Token ring hardware responds better than Ethernet hardware during periods of heavy use. |
Fiber optic |
Fiber-optic hardware provides very high bandwidth, but is usually too expensive to use throughout an entire network. |
Separate network |
Use a separate network to handle traffic between the highest volume workstations and Adaptive Server. |
Network overload – network managers rarely detect problems before database users start complaining to their system administrator.
Be prepared to provide local network managers with predicted or actual network requirements when they are considering adding resources. Also, monitor the network and try to anticipate problems that result from newly added equipment or application requirements.