Abstract-Cloud computing resources scheduling is essential for executing workflows in the cloud platform because it relates to both the execution time and execution costs. In solving the problem of optimizing the execution costs while meeting deadline constraints, we developed an efficient approach based on ant colony system (ACS). For scheduling T tasks on R resources, an ant in ACS represents a solution with T dimensions, with each dimension being a task and the value of each dimension being an integer ranges in [1, R] to indicate scheduling the task on which resource. With such solution encoding, the ant in ACS constructs a solution in T steps, with each step optimally selecting one resource from the R resources, according to both the pheromone and heuristic information. Therefore, the solution encoding is very simple and straight to reflect the mapping relation of tasks and resources. Moreover, the solution construct process is very natural to find optimal solution based on the encoding scheme. We have conducted extensive experiments based on workflows with various scales and various cloud resources. We compare the results with those of particle swarm optimization (PSO) and dynamic objective genetic algorithm (DOGA) approaches. The experimental results show that ACS is able to find better solutions with a lower cost than both PSO and DOGA do on various scheduling scales and deadline conditions.

Abstract-Machine learning (ML) approach to modeling and predicting real-world dynamic system behaviours has received widespread research interest. While ML capability in approximating any nonlinear or complex system is promising, it is often a black-box approach, which lacks the physical meanings of the actual system structure and its parameters, as well as their impacts on the system. This paper establishes a model to provide explanation on how system parameters affect its output(s), as such knowledge would lead to potential useful, interesting and novel information. The paper builds on our previous work in ML, and also combines an evolutionary artificial neural networks with sensitivity analysis to extract and validate key factors affecting the cloud data center energy performance. This provides an opportunity for software analysts to design and develop energyaware applications and for Hadoop administrator to optimize the Hadoop infrastructure by having Big Data partitioned in bigger chunks and shortening the time to complete MapReduce jobs.

Abstract-The success of Hadoop, an open-source framework for massively parallel and distributed computing, is expected to drive energy consumption of cloud data centers to new highs as service providers continue to add new infrastructure, services and capabilities to meet the market demands. While current research on data center airflow management, HVAC (Heating, Ventilation and Air Conditioning) system design, workload distribution and optimization, and energy efficient computing hardware and software are all contributing to improved energy efficiency, energy forecast in cloud computing remains a challenge. This paper reports an evolutionary computation based modeling and forecasting approach to this problem. In particular, an evolutionary neural network is developed and structurally optimized to forecast the energy load of a cloud data center. The results, both in terms of forecasting speed and accuracy, suggest that the evolutionary neural network approach to energy consumption forecasting for cloud computing is highly promising.

No abstract