Data Replication in a Distributed Heterogeneous Database Environment: An Open System Approach

This paper describes an open system approach used by the Sybase replication server for maintaining loose consistency of replicated data in a distributed heterogeneous database environment. In large distributed database systems, the same data are often replicated at multiple sites for quick data access and for high data availability. Because of the high cost of maintaining tight consistency of replicated data, most database systems choose to support loose consistency of replicated dam. The Sybase replication server is a transaction log based database replication system that allows row level data replication. It continuously propagates update transactions committed on the primary copy of data to the replicate copies stored in other databases. Unlike other data replication systems which are developed for specific database systems, the replication server is designed to be a separate process indcpendent of data servers which store the replicated dam. It provides an open system interface to communicate with heterogeneous data servers. The open system interface consists of two parts: the log transfer interface for accepting the log records for transactions executed on the primary copy of data, and the function string interface for users to translate commands in replicated transactions before they are applied to the databases containing replicate copies of data. The open log transfer interface makes it possible to replicate data from any databases that provide an interface to access their transaction logs. The open function string interface makes it easier to replicate data to databases through simple passthrough gateways. It also provides the flexibility for users to customize how transactions should be applied. We believe the replication server provides an important step towards providing practical solutions to data replication in large distributed database environments, most of which have data in heterogeneous databases.