DbHAaaS: Database High Availability as a Service

Database (DB) system applies "Undo" and/or "Redo" operations to recover the DB from a failure. The system uses "Checkpoint" operation to reduce the overall recovery time. The architecture of the cloud makes it possible to eliminate the effect of system failure by always taking the DB to the next consistent state. We present a novel scheme of eliminating the effect of such failure. Thus, in our system a transaction always commits irrespective of system failure. We refer to our system as AAP (Always Ahead Processing). We have made transaction execution persistent by combining together transaction execution and transaction failure and its subsequent recovery as one seamless operation eliminating system downtime during recovery. The end result is a DB system with high availability and fault-tolerance. Our work enables cloud providers to offer transactional HA-DBMS as an option that too with multiple data sources not necessarily relational.