Completed on 6 January 2006 by Christian Storm.

Elaboration

• Prof. Dr.-Ing. Oliver Theel

• Replication

Data replication is used to minimise the possibility of access failures to urgently required data (high access availability), but also to reduce access times to this data. The department has created a CORBA-based prototype for consistent data replication, which is used to manage replicated WWW documents. This prototype includes a component that allows any so-called static Coterie-based replication procedure to be specified and used to manage documents on the WWW. The specification of a desired special replication strategy is carried out by specifying an acyclic graph (in text form according to a formal language). The extension of this component to capture and process a further class of Coterie-based replication strategies has also been implemented: the extension relates to the so-called dynamic replication strategies. These are generally characterised by higher access costs on the one hand and higher access availability of the replicated WWW document on the other. What both approaches have in common, however, is that they assume a fixed number or upper limit of replicas at the time of design. This approach offers some advantages, but also has the disadvantage of not being able to increase the number of replicas if this is actually necessary due to the current user profile. Replication strategies that support a dynamic number of replicas, i.e. that have two operations a la create_replica() or delete_replica() in addition to reading and writing the replicated data, which consistently create a further replica at runtime or consistently delete it at runtime, could provide a remedy here. Consistent in this context means that all other operations continue to work correctly.

The task to be solved in this thesis is the conception of the operations create_replica() and delete_replica() for a general application to static, quorum-based replication strategies and their prototypical implementation in our research prototypes described above for the management of replicated WWW documents. By means of analytical as well as experimental investigations, the benefit and the effort of the two operations on the runtime costs of a replication strategy will then be demonstrated and evaluated both qualitatively and quantitatively.

• Knowledge of C/C++ as well as operating systems and distributed systems is desirable