A general stochastic model for dynamic locking in database systems

We present a novel stochastic model to study the performances of the two-phase dynamic locking in database systems with no-waiting policy. It is a general stochastic model to describe the database environment and transaction states in detail. It deals with the nonuniform access, write-locking, read-locking, and multiple transaction classes in a unique way. In the analysis, we first solve the steady-state probability of the system. Then, we give the mean number of transactions with k locks, the mean total number of locks held by all transactions, the mean number of data granules locked by a transaction, the mean number of writelocks and readlocks held by a transaction, and the mean number of locked data granules in a database. These parameters provide more insight into the detailed behavior of transactions and database systems. Finally, we calculate the system throughput and restart rate, which are the two principal performance measures.