Interact network considerations

A production installation of Interact spans at least two machines. In a high-volume production environment, with several Interact runtime servers and distributed databases, your installation might span dozens of machines. For best performance, there are several network topology requirements to consider.

*
executeBatch(startSession, getOffers, postEvent, endSession)
you do not need to enable session persistence (sticky sessions) between the load balancer and the Interact runtime servers. You can configure the Interact runtime servers session management for local cache type.
*
startSession
. . .
executeBatch(getOffers, postEvent)
. . .
endSession
and you are using a load balancer for your Interact runtime servers, you should enable some type of persistence for the load balancer (also known as sticky sessions). If that is not possible, or if you are not using a load balancer, configure the Interact servers session management for a distributed cacheType. If you are using a distributed cache, all the Interact runtime servers must be able to communicate via multicast. You may need to tune your network so that the communication between Interact servers using the same multicast IP address and port does not impede system performance. A load balancer with sticky sessions has better performance than using a distributed cache.
*
If you have multiple server groups using a distributed cacheType, each should use a unique multicast port. Using both a unique multicast port and address for each server group is better.
*
*
*
*

In a testing or staging installation, you can install Interact design time and runtime on the same machine. This scenario is not recommended for a production environment.



IBM Unica Interact
 
8.5.0
For more information, see our support and community site: Customer Central