Limited Preemptive Disk Scheduling for Real Time Database System

Abstract

Previously disk IOs have been thought of as non-preemptible operations. But in case of applications requiring short, interactive responses this non-preemption can be a disruption. One such domain is that of real-time disk scheduling. Such a blocking due to non-preemption is undesirable since it degrades the schedulability of real-time transactions as well as causes an hindrance to an execution of higher priority transactions. On the other hand making disk IOs preemptible produces many unnecessary preemptions. Arbitrary preemptions can introduce a significant runtime overhead and may cause high fluctuations in transaction execution times, so degrading system predictability. So there is a need of a viable alternative between these two extremes that are non-preemptive and fully preemptive scheduling models. As we know, before giving service to transactions, it is necessary to schedule them and find the required time. So, in order to know what this time required is and how to minimize it, it is mandatory to study the different parameters required for real time disk scheduling. This task can be achieved with the help of a mathematical model which shows how scheduling result of any algorithm can be evaluated. This paper derives a limited-preemption real-time scheduling algorithm and using a mathematical model it is evaluated and compared with the fully preemptive and non-preemptive scheduling models and shows how a limited preemption model acts as a viable alternative between these two extreme cases of fully preemptive and non-preemptive scheduling.