InfoCenter Home >
7: Multimachine management >
7.2 Managing workloads >
7.2.4 Using models and clones >
7.2.4.5 Server selection policies and transaction affinity

7.2.4.5 Server selection policies and transaction affinity

When you are cloning an application server, you need to take the following things into account:

Application server models and clones

A servlet engine is a server process that works with your Web server to handle requests for servlets and Web resources such as HTML, JavaServer Pages (JSP) files, and such. A server object, such as an application server or servlet engine, can be used as a model for creating multiple clones of that server. The clones are defined by a model. Changes to the model are propagated to the clones when the server clones are restarted.

The clones remain basically identical to the model, allowing work to be distributed to any one of them. Server selection policies determine how clients choose server instances within the group.

Server selection policies

The workload management server selection policy defines how clients choose among application server clones (instances). Select among these policies:

  • Random
  • Round-robin
  • Random prefer local
  • Round-robin prefer local

See article 6.6.22.0, for a detailed description of the server selection policies.

Transaction affinity for application servers

Regardless of the selection policy used, the workload management service attempts to choose an application server clone based on transaction affinity. Within a transaction, the first time a server is picked, the prevailing selection policy for the server group is applied. After a server is selected, it remains bound for the duration of the transaction.

For example, suppose the round-robin policy is specified for server group A with two application server clones, S1 and S2. A client has two concurrent threads, t1 and t2, with transaction contexts T1 and T2, respectively. Assume that thread t1 is first and needs to select a server from server group A; clone S2 is randomly chosen. When t2 tries to select a server from server group A, S1 is chosen based on the round-robin policy in effect for the server group. Subsequent requests to server group A are serviced by S2 for t1 and S1 for t2, based on transaction affinity.

Go to previous article: Containment relationships Go to next article: Security for cloned resources

 

 
Go to previous article: Containment relationships Go to next article: Security for cloned resources