Locking with prevention of cyclic and infinite restarting in distributed database systems

A new solution to the cyclic restarting and infinite restarting problems for locking schemes in Distributed Database Systems (DDBSE) is presented. The solution proposed is based on the data marking mechanism, which ensures the completion of each transaction in the system. The solution is fully distributed. It only requires information locally accessible on each site of the DDBS, and it intervenes into transaction processing only in the case of real danger of cyclic and/or infinite restarting. Simulation has shown that this solution significantly reduces the number of transaction restarts in DDBSs using locking schemes, and thus considerably improves DDBS performance.