Abstract:
Most of the companies distribute their
information per branch, ensuring that the data are
mainly updated locally, by the branch where they
were created. Examples of such companies could
be hypermarkets and banks. So, it seems advisable
to replicate all data objects, independent of the
company branch where they have been created. If
data are replicated, “remote” accesses can be
accomplished locally, improving access times as
well as availability, in case of node failures in the
network. However, each time a data object is
modified, updates have to be multicast to a given
number of database replicas.Data replication
among different sites is viewed as a way to
increase application performance and its data
availability.
This system proposes a concurrency
control and recovery in a middleware architecture
called COPLA (Common Object Programmer
Library Access). This architecture provides
persistent object state replication. This system is
based on time-stamp, it is an adaptation of the
Optimistic protocol to this architecture. The
recovery process of this system allows
applications to continue (or start) executing
transactions at all nodes, even in the node being
recovered.
