ON THE MANAGEMENT OF LONG-LIVING TRANSACTIONS

Long-living transactions are those transactions that exist in the system for relatively long periods of time. Because of their special nature, these transactions cause scheduling difficulties. Traditional concurrency control methods are unsatisfactory for their management. In this paper, we propose methods to handle long-living transactions efficiently. Our algorithm extends some of the previous works on this problem in several important ways. We first propose several performance criteria to be satisfied by any solution and then provide an efficient solution for the management of long-living transactions. In this process, we identify several aspects to the problem and suggest approaches to these issues. Extensions to the case of a multitude of long transactions and distributed data base systems are considered. We show that attractive solutions are possible in these situations. © 1990.