Content area
Full text
Abstract- Techniques for estimating and predicting the availability of CPU resources for a task while executing in a singleand multi-core systems are introduced. These predictions have a significant effect in areas such as dynamic load balancing, scalability analysis, distributed task queue scheduling, and others. Analytical models proposed in this paper is useful for predicting the availability of CPU resources while executing a new task. The model can predict the allocation of CPU for a new task without prior knowledge of the run queue and system state. To validate the introduced prediction models, a dynamic monitoring tool (simulator) is developed. This real-time monitoring utility is responsible for measuring states and availability of resources based upon receiving resource request from client program. Extensive experimental studies with real-world benchmark programs and statistical analysis are performed to measure the accuracy of models. The performance of introduced monitor tool is evaluated as well while extracting resource availability data from the computer grid.
Keywords- CPU availability; Computer grid; CPU Load; Distributed system; Multi-core processors; Performance Prediction;
(ProQuest: ... denotes formulae omitted.)
I. Introduction
Correlation refers to any of a broad class of statistical relationships involving dependence. Correlations are useful because they can indicate a predictive relationship that can be exploited in practice. The approach of proposing the new model in this paper is correlation based empirical model using time series analysis. Data points are measured typically at successive time instants because time series analysis assumes that sequential values in the data file that embodies repeated quantities are taken at equally spaced time intervals [3]. A sequence of methods for analyzing time series data in order to extract meaningful statistics and other characteristics of the data has been employed in this paper. The introduced mathematical models are based on an empirical average-case study in two different extreme situations, when the CPU is lightly and heavily loaded. Additionally, the scheduling pattern of tasks was observed for a long period of time to incorporate the accuracy.
When the number of threads assigned to a multi-core processor is less than or equal to the number of CPU cores associated with the processor, then the performance of the CPU is predictable, and is often nearly ideal. When the number of assigned threads is more than...