A Dependable Computing in Database for Managing Mass Connection Systems

Most operations in database (DB) is based on the deferred update replication technique, especially in distributed environments. In deferred update replication, transactions are executed on a single host and broadcast to other hosts at commit time. Upon delivering a transaction, each host confirms it to ensure a globally serializable execution. The mechanism is subject to CAP theorem in order to ensure data consistency and availability. This paper proposes a dependable cache query mechanism in distributed relational database (DRDB). The key insight of the cache query mechanism provides high performance, strong consistency and availability for mass connection systems. In addition to presenting the cache query mechanism, we elaborate its implementation, and provide an experiment and analysis of its performance.