摘要
Disaster recovery （DR） techniques ensure the data safety and service continuity under different natural and human-made disasters by constructing a high reliable storage system. Traditional disaster recovery methods are structure-dependent. It is hard to share the DR resources between different DR systems, which made it expensive. We present a structure-independent disaster recovery theory and its implementation methods in this paper. By backup the whole system but not just the data, the goal of device and application-independent disaster recovery has been achieved. We further present a parallel recovery model and an on demand data retrieval method based on the theory. Some implementation details of prototype recovery system are also discussed. With the methods independent from specific devices or applications, the cost of disaster recovery infrastructure can be essentially reduced by resource sharing. Experiments show that the recovery time has also been greatly shortened with little service degradation.
Disaster recovery （DR） techniques ensure the data safety and service continuity under different natural and human-made disasters by constructing a high reliable storage system. Traditional disaster recovery methods are structure-dependent. It is hard to share the DR resources between different DR systems, which made it expensive. We present a structure-independent disaster recovery theory and its implementation methods in this paper. By backup the whole system but not just the data, the goal of device and application-independent disaster recovery has been achieved. We further present a parallel recovery model and an on demand data retrieval method based on the theory. Some implementation details of prototype recovery system are also discussed. With the methods independent from specific devices or applications, the cost of disaster recovery infrastructure can be essentially reduced by resource sharing. Experiments show that the recovery time has also been greatly shortened with little service degradation.
基金
Supported by the National Basic Research Program of China (Grant No. 2007CB311100)
