Deadlock prevention and avoidance methods using timestamps for locking control in distributed database systems

Abstract Locking is used widely for concurrency control in database systems, and locking methods are categorized as static locking or dynamic locking. For distributed database systems, this paper proposes a new deadlock prevention method for static locking and a new deadlock avoidance method for dynamic locking. The common advantages of these methods are as follows: (1) system performance is improved by using time stamps; (2) since transactions are executed without explicitly using time‐stamp ordering, after a transaction lock is permitted, even when another transaction with a smaller time stamp arrives, the rollback of the transaction is unnecessary; and (3) since time stamping is used only for deadlock resolution, the proposed methods can easily be introduced to existing distributed database systems that use locking control.