Adaptive replication strategies and software architectures for peer-to-peer systems

The use of replication techniques for data in distributed information systems aims at improving non-functional properties of these systems. Particularly, availability, reliability, and performance should be increased. When considering heterogeneous, autonomy-preserving systems such as distributed digital libraries, the replication techniques employed should be able to adapt to varying properties of the individual systems at run-time. Adaptive replication strategies need to be realized by a software architecture which provides the context for their implementation. Digital library systems whose information providers are organizationally closely related, for example under the umbrella of a single institution, should be tightly coupled, in order to allow for a high degree of consistency to be maintained. Other digital libraries, such as those of separate publishers, should be coupled less tightly, to allow for retaining their autonomy. This can be realized by a replication strategy based on peer-to-peer (P2P) techniques which combines intra-institutional and interinstitutional replication strategies. The overall replication strategy is evaluated and optimized at run-time, such that domain-specific requirements are incorporated into a multi-dimensional problem model.