Session management in clusters is a serious problem faced by most enterprises that provide interactive Web applications. Existing session persistence techniques, such as centralized database persistence, known as Microsoft SQL server or memory-to-memory replication, known as Broadcasting algorithm, suffer from severe limitations. Centralized database persistence may suffer from Single Point of Failure while the Broadcasting algorithm may suffer scalability problem since the replication of session objects adds significant overhead. The Fetch on Demand algorithm is introduced as a software solution that overcomes the above-mentioned challenges associated with session management. The study focuses on implementation of the Fetch on Demand algorithm, analyzing the performance and comparison to existing designs.