How to stabilize Informatica domain in a non optimal environment

There are four types of settings that influence performance, and can help stabilize a domain:

1. Resilience: Settings that influence the keeping alive of processes during (temporary) network failures
2. Timeouts: Values that govern if or when a connection is lost
3. Intervals: Values that govern how often to check whether a process or service is alive
4. Performance

Resilience

Resilience is the ability of PowerCenter service clients to tolerate temporary network failures until the timeout period expires or you resolve the system failure. Clients that are resilient to a temporary failure can maintain connection to a service for the duration of the timeout.

There are two variables associated with resilience:

Limit on Resilience Timeouts: 180 (default). The amount of time a service waits for a client to connect or reconnect to the service. This limit can override the client resilience timeouts configured for a connecting client. This is available for the domain and Application Services.

Resilience Timeout: 30 (Domain), and 180 (Repository and Integration Services). The amount of time a client attempts to connect or reconnect to a service. A limit on resilience timeouts can override the timeout.
In general, it is advisable to keep all limits at the same level, to keep the interdependency even.

Timeouts
Connection Timeout: 40000 (server/tomcat/conf/server.xml). The number of milliseconds the Connector (https) waits, after accepting a connection, for the request URI line to be presented. This parameter is a general tomcat server parameter and is to do with how long the server leaves a connection between the browser and server open after a response is complete.
Repository Service: Database Properties: DatabaseConnectionTimeout: 180 Period of time that the Repository Service attempts to establish or reestablish a connection to the database system. Default is 180 seconds.

Refresh intervals
Domain: Properties: Custom Properties: MasterDBRefreshInterval: 8 (effective 28 seconds with factor 3.5). This is the interval the domain master gateway uses to write its entry in the domain database, table PCSF_MASTER_ELECTION, determining it is alive. If this update fails four consecutive times, the master gateway node shuts down. If there are multiple gateway nodes in the domain, a new master gateway will be elected, assuming the domain database is alive.
/server/tomcat/bin/infaservice.sh -Dinfa.masterUpdateTimeInterval=15000.
The domain service marks a node inactive if the master gateway node does not gets a notification from the node within six times 15 seconds. When the CPU load on the machine is high, the worker node cannot send the notification in time. The master gateway node will mark the node as inactive and all the services on the node will not be available, despite the node is actually up and running. Change the master gateway timeout by adding the following system property to INFA_JAVA_OPTS: setenv INFA_JAVA_OPTS="${INFA_JAVA_OPTS} -Dinfa.masterUpdateTimeInterval=X"
Repository Service: Properties: Advanced Properties: HeartBeatInterval: 60 Interval at which the Repository Service verifies its connections with clients of the service.

Performance
Repository Service: Properties: Database properties: Optimize Database Schema: No. Enables optimization of repository database schema when you create repository contents or back up and restore an IBM DB2 or Microsoft SQL Server repository. When you enable this option, the Repository Service creates repository tables using Varchar(2000) columns instead of CLOB columns  wherever possible. Using Varchar columns improves repository performance because it reduces disk input and output and because the database buffer cache can cache Varchar columns.

To use this option, the repository database must meet the following page size requirements:

IBM DB2: Database page size 4 KB or greater. At least one temporary tablespace with page size 16 KB or greater.
Microsoft SQL Server: Database page size 8 KB or greater.