Research On Prediction Of Timing Parameters Based On History Trace

With rapid development of computer network, the applications of distributed real-time systems become more and more popular. For the efficient management of system resources to meet the real-time requirement and predictability of real-time system, we need predict the timing parameters of real-time tasks. The timing parameters show time constraint and behavior of real-time tasks, which predicting real-time task execution time is very important and significant. Predictability of real-time task execution time is an important content in real-time system research, and it is closely related to and affect tasks scheduling and resource allocation. Linear time series as a statistical probability method has been used in various fields, for predicting objects changes over time, time series analysis is the most basic method in predicting system. This article explains how to get the timing parameters by using time series analysis method to study the host load history trace, and then predict the task execution time base on host load.This paper introduces the concept of the real-time system, timing parameters related terms, the definition, access and characteristics of host load, also introduces RT-CORBA technologies; then elaborates the classification of analysis methods, detailedly analyses time series and regression analysis prediction methods, and prediction method of assessment; then, analyses the predictability of real-time systems, the predictability of the execution of the tasks, introduces detailedly the process of constructing time series model by research on the host load history trace: pretreatment, smooth processing, pattern recognition, model-order determination, parameter estimation, prediction of data series, and detailedly analyses the factors of predicting task execution time; and then, dicusses the relationship of host load and the task execution time, and prediction algorithm of real-time task execution time, explains experimental operating environment, and proposes a task execution time prediction model based on the host load, and gives a prediction analysis algorithm, analyses the impact of the prediction; finally, put algorithm of task execution time prediction base on host load history trace in distributed network environment for the conduct of the experiment, the predicting result based on the above is compared with the real result, the comparing result shows the rightness of the work, the result has the good reference significance for prediction of real-time task execution time based on host load.