From a performance perspective, you need to configure the Sterling Selling and Fulfillment Foundation system so that it can quickly discover the Oracle failure and to quickly recover the connections.
The Sterling Selling and Fulfillment Foundation system is made up of client programs that connect to the Oracle instance. These include:
The default time for a connection request to an unavailable node to timeout is deliberately set to a high value. This value allows connection requests (e.g., telnet connections) the opportunity to find the node on the Internet. This setting is less applicable in a high-speed switched network.
On Solaris, a telnet to a nonexistent node takes about 2.75 minutes to timeout. On HPUX11, the timeout is around 75 seconds.
The connection timeout value can be tuned down by issuing the following ndd commands:
ndd -set /dev/tcp tcp_ip_abort_cinterval 1000
ndd -set /dev/tcp tcp_rexmit_interval_initial 200
ndd -set /dev/tcp tcp_rexmit_interval_max 5000
For information about tuning TCP/IP, go to http://docs.oracle.com/cd/E19455-01/806-6779/6jfmsfr86/index.html.
Fast Application Notification (FAN) provides RAC the ability to inform the client programs the status of the cluster. With FAN, the client programs, especially those with connection pools, can drop stale connections to failed nodes.
Sterling Selling and Fulfillment Foundation does not support FAN.