Reconciling replication and transactions for the end-to-end reliability of CORBA applications

The CORBA standard now incorporates support for reliability through two distinct mechanisms – replication (using the Fault Tolerant CORBA standard) and transactions (using the CORBA Object Transaction Service). Transactions represent a roll-back reliability mechanism, and handle a fault by reverting to the last committed state, and by discarding operations that were in progress at the time of the fault. Replication represents a roll-forward reliability mechanism, and handles a fault by re-playing any operations that were in progress at another operational replica of the crashed server. Most of today's enterprise applications have a three-tier structure, with simple clients in the first tier, servers in the middle-tier to perform the processing, and databases in the third tier to store information. For such applications, replication is required to protect the middle-tier processing, while transactions are required to protect the third-tier data. This requires the reconciliation of roll-forward and roll-back reliability mechanisms in order to protect both data and processing, and to provide consistent end-to-end reliable operation. This paper looks at the issues of integrating replication with transactions for three-tier enterprise CORBA applications, with particular emphasis on reconciling the Fault Tolerant CORBA standard and the CORBA Object Transaction Service.