FTL Stores Overview
FTL stores function very similarly to file-based stores. When using FTL stores, all pending persistent message data and state information is maintained in both server memory and in an FTL server cluster. Keeping this information in memory reduces the amount of communication needed with FTL and facilitates faster message processing.
The EMS server must have access to all state information and pending messages before coming online. The in-memory design of FTL stores requires that the EMS server reads the contents of all stores into memory upon start-up or failover to achieve this.
The main differentiating factor between FTL stores and file-based stores lies in their underlying storage solutions and the implications of those for fault-tolerance. To run a fault-tolerant EMS server pair with file-based stores, you will need a shared storage mechanism that meets the Support Criteria for shared state. FTL stores on the other hand, rely on the persistence and replication features of FTL for their backend and require the deployment of an FTL server cluster. When making the choice between the two store types, in most cases the primary deciding factor should be whether you have access to a shared storage mechanism or not.
Communicating with TIBCO FTL for message processing involves more overhead than the simple file reads and writes performed by file-based stores. Performance may vary depending on the usage and environment, but in general FTL stores are likely to be slower than file-based stores due to this.