A study on parallel real-time transaction scheduling

With the rapid development of modern real-time applications, the need of high scalable and predictable real-time transaction processing technology becomes more and more urgent. In this paper, we focus on realtime transaction scheduling algorithm in shared-nothing parallel database systems. We propose and evaluate a new time-stamp based scheduling protocol, which uses priority-based time-stamp to implement parallel sub-transactions synchronization. The experimental results show our new protocol can better resolve the conflict between the synchronization controlling and the communication overhead. Therefore the protocol performs well in the cases when the system overload is heavy or the skew problem is serious.