Data in a replicate database is “loosely consistent” with the data in the primary database. Replicate data lags behind primary data by the amount of time it takes to distribute updates from the primary database to another part of the replication environment. This latency can be measured in seconds (or less) when the system is working properly. If a component fails—if, for example, a network connection is temporarily lost—updates can be delayed for minutes, hours, or days. Thus, latency information can be used to monitor the performance and health of the replication environment.
Although replicate data may lag behind primary data, it is transactionally consistent with the primary data. Replication Server delivers transactions to replicate databases in the order they are committed in the primary database. This ensures that the replicate data goes through the same series of states as the primary data.
The importance of loose consistency varies by application and even within an application. Some applications tolerate average system lag time and occasional, longer delays with no special provisions. Some require special handling when the lag time becomes too great, and some require special handling for certain types of transactions.
