Abstract: Successful submission before deadline is the key ingredient for applications in Real-Time DataBase Management Systems（RTDBMS），while traditional DBMS focuses main on the opposite goal of high throughput of the system.A two-Phrase Speculative Concurrency Control approach based on the deadline（PSCC） is therefore addressed to raise the successful submission for real-time transactions under stringent time constraints.By achieving a tradeoff between the optimistic concurrency control and the passive concurrency control，the PSCC can speculate on the suspicious conflicts according to their deadline，restart transactions with more possibility to imperil the DB consistency while allowing those to execute continually with less possibility.At the verification phrase of the transaction，the broadcasting approach is used to guarantee the DB consistency as the OCC-BC does，but those transactions violating submission consistency rule will be temporarily blocked and a priority escalation approach is used to avoid the restarting of blocked transactions.Finally，the algorithms of PSCC are designed.Performance evaluations show that by using PSCC algorithms，the missing rate and restarting probability can be reduced than using OCC-BC and HPA.

Key words: real-time database systems, concurrency control, speculative concurrency control, priority escalation

摘要： 在实时数据库应用环境中，最大的并发数或者吞吐量不再作为性能的重要衡量尺度，而把在截止期之前完成的事务数量作为决定性的性能指标。因此，综合乐观并发控制（OCC）和悲观并发控制（PCC）的方法，提出了基于截止期的两阶段可推测并发控制技术（PSCC）以提高事务满足时间限制的可能性。首先，基于事务的截止期对潜在的可能带来数据不一致性的威胁进行推测，尽早地重启那些变为现实的威胁可能性较大的冲突事务，允许变为现实可能性较小的冲突事务继续执行。其次，在校验阶段采用广播式的提交技术控制未违背提交一致性的事务的提交，保证了数据库一致性，同时暂时阻塞违背提交一致性的事务并利用优先级升级技术减少被阻塞事务的不必要的重启。最后，给出了PSCC的行为规则及算法描述。性能测试结果分析表明，PSCC比OCC-BC和HPA有更低的失败率的重启率。

关键词: 实时数据库, 并发控制, 可推测并发控制, 优先级升级