Research On Acquire Technology Of Real-time Executable Probability Facing Streaming Media Application

With the increasing complexity of network application types and modes of application of the gradual maturity, the implementation of real-time tasks on the Internet has become the development of the times demand. But, two conditions need be satisfied in order to execute the real-time tasks. These are, the necessary condition, which the supply of network resources must exceed the demand, the sufficient condition, which the execute time must satisfy the time limit. This and stochastic changes in network resources form a pair of prominent contradictions. How to accurately obtain the real-time executable probability of real-time tasks and carry out forecasting with great accuracy, is the most urgent problem to be solved.The research of this paper has practical significance to the admission control of real-time tasks, reasonable assignment of shared network resource, and increasing the utilization ratio of resource.The existing methods and tools can not acquire and forecast real-time executable probability directly. Setting in the soft real-time application of streaming media under the Linux and beginning with obtained the status of network resources,this paper puts forward a method of acquiring the network resources margin executable probability , in order that we can get the time limitation curves of network resources margin executable probability . In the meantime, because the existing methods and tools can not forecast the executable probability of real-time tasks fully and accurately ,this paper puts forward a long- and medium-term and realtime hybrid forecast method to forecast the resources margin executable probability. The experimental results show that the forecast methods of this paper puts forward have a higher forecast accuracy.The forecast curves and the real curves are approaching on the values and trends,and the average forecast error is less than 8.5%.The work of this paper shows a beneficial exploration on key issues on practise of the probability real-time theory of the distributed interactive environment.Main tasks include:1.According to the special requests to the network resources of this topic and the defects of existing acquiring techniques,this paper puts forward a technique to realtime acquire the network resources margin.The technique acquires the three types of network resources(Computing Resource,Memory Resource and Traffic Resource) margin realtime by programming base on Linux/proc file system. Contrasting with the existing tools or memory modules access, the method of this paper put forward has fewer costs, more realtime and more information.2. According to the current situation of resources margin, this paper puts forward a method to acquire the time limitation curves of resource margin probability. Based on the sampling reports of resource margin,the method figures out the distribution executable probability of network resources margin. Using the data in different time periods ,this paper carries out integral evaluation, and gets the 24 hours time limitation map of resource margin executable probability finally.3. About the forecast of network resources margin executable probability,this paper puts forward a long- and medium-term and realtime hybrid forecast method.This method organic combin the long- and medium-term and real-time forecast methods'advantages. The long- and medium-term forcast use a combined model constituted by the multiplication and winter linear model.The realtime forcast use a method named self-adapting algorithm based on forecast irrelevance threshold values.It can overcomes the defects of the long- and medium-term and real-time forecast method .This method can eliminate the seasonal element of multiplication model in effect, accurately grasp the periodicity regular and the seasonal trend of resources margin executable probability, as well as can real-time forecast the fluctuation of probability with less spending within a short period of time.4.This paper sets up the experiment platform of streaming media application, implements and verifis the method put forward by this paper to acquire and forecast executable probability of network resources margin.