A pure lazy technique for scalable transaction processing in replicated databases

Recently, there have been proposals for scaling-up a database system using lazy replication. In these proposals, system scale-up is achieved through the addition of secondary sites which hold replicas of the database at a primary site. The addition of more secondary sites improves system performance by allowing read-only transactions to be serviced at the secondary sites. However, the scalability of the distributed system is limited by the single primary site which services the update workload. As the workload scales-up, an increasing update load is placed on the primary site, which suffers from performance degradation. We address this problem by proposing a pure lazy solution for scaling-up the database system. Our techniques provide scalability and avoid transaction inversions, which can occur when transactions access stale replicas.