Research On The Executable Probability Of Real-Time Task In The Shared Network

Recently, real-time tasks are applied throughout every corner of the Internet, along with the quality of service requirements raised. Admission control becomes more important as an part of Service quality management. This paper investigates the influence of executable probability on real-time admission control in the streaming media service field, which could provide a more reliable basis for the management of real-time systems and the scheduling of real-time task.Firstly, the resource consumption features of real-time tasks are analyzed. As the resources in a shared network are always limited, getting the system load accurately was a prerequisite for admission control. A limited variable time window measurement method is designed in this paper, which can obtain the load of multi-resource dynamically and provide a more comprehensive foundation for the admission control.Secondly, this paper uses grey theory to predict the executable probability, as the current algorithms in the resources forecasting field has some deficiencies. An improved grey predict model with testing and error correction methods is designed in this paper. Thirdly, this paper presents the Executable Probability Based Admission Control method with Feedback Mechanism (EPBAC-FM). The method has been programmed in the Linux system by the author.Finally, some tests are implemented to check the feasibility of the EPBAC-FM.The results shows that: the average relative error of the improved grey predict model was 4.41%, far better than the traditional gray prediction theory. The executable probability shows a high credibility for real-time tasks admission control. EPBAC-FM could make effective decisions depending on the system load, and the packet loss rate is reduced by half in high load conditions, comparing to traditional measurement based admission control (MBAC).