Performance Evaluation on a Real-Time Database

We have implemented an object-oriented real-time database system called BeeHive. Using BeeHive, the performance of two data-deadline cognizant scheduling policies, called EDDF and EDF-DC, and a baseline EDF policy all with/without admission control are evaluated through extensive experiments. We identify the ranges where data-deadline cognizant scheduling policies are effective. We also show that when admission control is used it plays a more significant role than specialized real-time data cognizant policies. Our admission control policy considers not only CPU utilizations, but also I/O requirements and utilizations. These results represent one of the few sets of experimental results found for an implemented real-time database; almost all previous results are via simulation.