Introduction of the new Operating System Kernel Internals for the New Metrics for the Performance Prediction on the Clouds

Abstract

In this paper, the introduction of the new OS kernel internals for the new metrics for the grid and cloud performance prediction is explained. This new introduction is named as Division of Load (DOL). The DOL method breaks down the CPU load by individual users and then separates the Disk IO load from the CPU load. In the first part, the concepts of the load signals are shown theoretically and experimentally as these metrics are introduced into the kernel for the first time. In the second part, the required code changes which are introduced to the OS kernel are discussed. The separation will help to collect the computer loads separately for individual users as CPU loads and Disk IOs loads. Such a move will open Grid, Cluster and Cloud Performance Predictors to use the divided data archives for better predictability of both CPU and Disk IO loads. Many existing Grid and Cloud resource prediction engines are going to be advantaged by this data purification and specialization.