Performance evaluation of a high-value scheduling policy for real-time main memory databases

With the decrease in hardware prices, storing databases in main memory has become feasible for meeting the high-performance requirements of real-time applications. In this paper, an efficient scheduling policy is proposed for scheduling transactions with values in real-time main memory databases, where the value of a transaction means its worthiness to a real-time application if it is completed before the deadline. Through detailed performance evaluation, the proposed policy is shown to outperform the existing policies in both minimizing the number of missed transactions and maximizing the obtained values under various system environments. A cost formula for evaluating the optimal memory size to store partial databases such as to maximize the net profits is developed from the economic standpoint. Some suggestions are given for scheduling value-based transactions in real-time main memory databases.