Key Technologies Of Green Cloud Computing Supporting Agile Service Optimization

The rapid development of cloud computing has caused large-scale expansion of cloud data centers,which leads to huge amount of energy consumption.The energy problem has attracted the attention of the academic and industry.How to achieve green cloud computing has been a key issue for the sustainable development of cloud computing.Besides,new models such as mobile cloud computing and cross-cloud service composition have been emerged.The features such as instability,real-time and request diversification have also brought challenges for supporting agile service optimization in green cloud computing.The service optimization in cloud computing environment can be carried out from cloud service provider and the infrastructure provider,which respectively correspond to service composition optimization techniques and virtual resource scheduling techniques.Although service optimization in green cloud computing has achieved some works.Current research works of service scheduling in cloud environment ignore the research of problems such as demands for agile services,multi-objective optimized resource allocation,and diverse resource requests.As a result,agile service optimization in green cloud computing still faces challenges as follows:1)Traditional service optimization methods are difficult to adapt to the dynamic environment quickly.The failure problem of the service optimization process may be caused,which not only leads to the waste of computing resources but also affects user experience.So the agile service optimization mechanism is needed.2)Cloud infrastructure service providers usually favour the energy aware resource provisioning strategies,which will affect the agility requirements of real-time tasks.So the consideration of trade-off in energy consumption optimization and response time is needed.3)There are applications with different resource request features in cloud platform.So overall consideration in service scheduling according to different resource requirements is needed to meet the requirement of agility and support green cloud computing.In view of these challenges,the solutions for green cloud computing supporting agile service optimization are proposed,which are described as follows.1)To realize the green cloud scenario for supporting agile services,a green cloud computing framework is proposed to support agile service optimization.The framework consists of four levels,which are service resource level,service discovery level,agile service optimization level and green resource scheduling level.Specifically,the service resource level is responsible for providing various cloud services and cloud resources through distributed service brokers.The service discovery level is to collect the registration information of cloud services and provide all kinds of services that can meet the applications for the agile service optimization level.The agile service optimization level is to provide service composition plans for user applications based on the agile service optimization method.Finally,in green resource scheduling level,agile service scheduling methods are respectively provided for real-time tasks and hybrid tasks to meet the requirements of diverse tasks and reduce the energy consumption of cloud data centers.2)To meet the requirement of users for QoS(Quality of Service)assurance and agility,an agile service composition optimization method in cloud computing environment is proposed.Specifically,an agile service composition optimization model is designed for cross cloud service composition.And a fine-grained service selection method is provided to deal with changes in services caused by various uncertain factors.Moreover,service pairs are built by multi-dimensional QoS attribute aggregation functions,and SAW(Simple Additive Weighting)method is introduced to calculate the objective utility value of service pair.Once a component service is selected,it should be executed immediately.At the same time,the next component service with the optimal predicting performance is selected based on the objective utility function.Finally,service composition recovery mechanism is proposed for unexpected situation caused by the unstable environment in the process of service composition.The agile service optimization method can quickly react to the changes caused by the unstable environment with near-optimal performance.3)To meet the agility requirement of real-time tasks in energy-aware resource scheduling,a resource scheduling method for trade-off between energy and performance is proposed.Specifically,the energy-aware resource scheduling problem is modeled by incorporating the virtualization technology.Moreover,an energy-aware resource provisioning algorithm is developed for generating available allocation plans for real time task by exploiting the elasticity of virtual resources.Then task-resource mapping method based on multi-objective optimization is proposed to optimize resource provisioning and set utility function.Finally,the optimal resource scheduling solution can be determined by the utility value,which can not only reduce energy consumption for data centers but also meet the agile requirement of real-time tasks.4)To meet the requirement of energy consumption optimization and different types of task requests,a hybrid task scheduling method with deadline-constrained is proposed for green cloud computing.Specifically,the urgency level is defined to determine the scheduling priorities of the waiting tasks by considering whether the load of task is divisible firstly.Moreover,a resource scheduling method for indivisible tasks is proposed.The indivisible task will be scheduled to the resource with minimum energy increment.And a resource scheduling method for divisible tasks is proposed by exploiting the computing parallelism of the divisible tasks,in which scheduling solutions for divisible task can be generated dynamically based on the maximum energy consumption rate strategy.Finally,the hybrid task scheduling method can improve the energy consumption while meeting the deadline of tasks.