Özet:

Abstract Cloud services are generally seen as a promising technique developed to achieve the highest computation service needs. However, such high-performing level of computing services can lead to the highest level of failure rates owing to a wide range of components and host servers which are filled with intensive job scheduling problems. Therefore, failure which occurs in one component or sub-system will lead to the unavailability of the computation services for the system. In this research, we suggest a new effective model called adapting fault-tolerant model (AFTM) which aimed to examine the optimization of job scheduling problem in computing infrastructure based on Particle Swarm Optimization (PSO), Apache Sparka and Ant Colony Optimization (ACO). The proposed approach covers the implementation and analysis of virtualizations with the job task selection to health monitoring for fault diagnoses based on Apache Spark. The objective is to find the cost trade-off between the allocated memory and CPU execution time required by virtualization services created by the end-users. The evaluation of the empirical performance of the proposed approach results outperforms PSO algorithms and traditional Genetic Algorithm (GA) in terms of the allocated memory and the time of CPU execution.

Anahtar Kelimeler: