为解决大规模交易中复杂事务密集访问引起的关键事务调度性能低下问题,文中通过研究交易事务的分类特点和结构特征,提出了一种基于虚拟截止期和时间戳排序的双级调度策略,通过分割长撮合事务和确定合理的步长因子,保证关键事务的优先调度,模拟和测试结果表明,在不产生事务延迟的条件下,交易事务的处理速率为1970事务/秒,平均响应时间为0.5ms,满足大规模电子交易的需要.
To solve the problem of performance worsening on scheduling key transactions induced by intensive data accessing in large scale e-trading, the paper classifies the varieties of trading transactions and investigates on their unique structures, and presents a two-stage schedule policy based on virtual deadline and timestamp ordering. It guarantees the prior scheduling of primary transactions by adopting both methods of proper segmentation of long matchmaking transaction and rational selection of amending factor. The results of simulations and tests show that, trading transactions can be totally processed up to 1970 within one second and each close to 0. 5ms in no delay condition, meet the need of scheduling concurrent transactions in large scale e-trading.