A FRAMEWORK FOR PROCESS-RESOURCE REQUIREMENTS PREDICTION IN A FACTORED OPERATING SYSTEMS USING MACHINE LEARNING

Abstract

The evolution of multicore technology has come with new challenges to process scheduling. An intelligent process resource requirement prediction framework in a factored operating system (FOS) was developed using machine learning. To simulate this system, an array of positive integer numbers in the interval of 1 to 10, with a maximum length of 20, was used to represent process. The properties of this array (array length and sum of the array elements) were used as the parameters of the application processes. These served as input into the machine learning network to predict the process’ size and hence, the resource requirements for each process. The framework was simulated using Java 2EE. Experimental results of the framework showed that prediction of processes’ sizes with enhanced resource requirement allocation, and hence, required faster and efficient scheduling of processes to multiple resource cores, with an increased system throughout.