Research And Design Of Real-time Transaction Processing

Real-time database is widely used in the industrialcontrol,telecommunication, finiance area.The amount of data it operates is more and more.The process is more and more complex.The research in the real time database is mainly in the Storage management,concurrency control,recovery,etc.Because the memory cost is becoming lower and lower,it is used as the storage widely in the Real-time database.The author researches in the concurrency control,including the pessimismistic method and the optimistic method and compares those methods.Besides,the difference in the concurrency control between DRDB and MMDB is analyzed and the examples of the database transacion realization are illustrated.As the multi-processor is cheaper and cheaper,the performance enhancement from the memory storage aspect is not enough.The advatage of the parallel processing udner multi-processor can be applied.The article introduces the parallel processing method in the CacheDB,The separate action is divided into several tasks and executed in separate processor.As the execute time of the actions,the time limit of the real-time transaction can be satified .The real-time performance may rise.Besides,the cost of the parallel processing is analyzed.In the parallel processing,the race of the data access is more sieverity,including the processor and memory.In this article,the lock method is optimized,it is called the adaptive lock.It is based on the pessimistic concurrent control.The lock granuanity is decided by the cases of race.The data access conflict is monitored and stored.The more concurrent users,the low granuality the lock is.By this way the whole system is balanced by the performance and cost.In the end,a performance test is made,including the comparison of the parallel procesing and serial processing and the effect of the concurrency control mentioned above.