A CORBA Object-based Caching with Consistency

Abstract

For many distributed data intensive applications, the default remote invocation of CORBA objects to a server is not acceptable because of performance degradation. Caching can improve performance and scalability of such applications by increasing the locality of data. This paper proposes a caching approach that optimises the default remote invocation behaviour of CORBA clients. Efficient fine-grained access to remote objects requires objects to be shipped to clients and cached across transaction boundaries. This approach is based on cache consistency via backward validation, generic approach for cache storage, object based data shipping and replication management. These features are introduced without changing the object definitions that the client may already depend upon. An implementation of the proposed caching approach is done on Orbix by extending the smart proxies. We have also provided a test with different client workloads. The results demonstrated a significant performance increase, in terms of transactions per second.