Energy Sensitive And Failure-oriented Multi-cloud Resource Scheduling Method

Cloud computing can expand flexibly and provide required resources in accordance with user requirements.Computing power and storage have become commodities for service providers and users.Due to the rapid growth of the scale of multi-cloud environments and the growthof tasks and users,the number of heterogeneous distributed data centers(DDCs)in multi-cloud environments is increasing.In a multi-cloud environment,security,reliability,and task execution time constraints are very important for tasks submitted by users in different geographic locations.Task scheduling is a prerequisite for optimizing task execution performance and improving the utilization of heterogeneous resources in a multi-cloud environment.In a distributed data center environment,it is necessary to respond to tasks submitted by users from different geographical locations.Under different resource requirements,the appropriate data center should be selected for the task efficiently and reasonably be responsible for task execution.At the same time,tasks need to be assigned to suitable Internet Service Providers(ISPs)for task transmission,which is an urgent problem in a multi-cloud environment.With the increase of distributed data center workloads,green energy(such as solar energy,wind energy)and Energy Storage Devices(ESDs)are integrated into Distributed Micro Grids(DMGs).Providing energy for distributed data centers is a research direction that has attracted attentions in recent years.In this thesis,a detailed study of task scheduling and resource management in a multi-cloud environment is studied,and new scheduling strategies are proposed for different scheduling objects to form a complete scheduling method.The main work is as follows:For the task scheduling strategy in the case of failures,the goal is to optimize the energy consumption,execution cost,and task rejection rate of the task under the constraints of reliability and task execution delay.First,for heterogeneous data centers distributed in different geographic locations,a scheduling strategy between tasks and data centers that dynamically adjusts and optimizes the target preference weights is proposed.Then,according to the task scheduling strategy mentioned above,considering the random occurrence of machine failures in the data center and the fluctuation of related resources,according to the set impact indicators and thresholds,the proportion of reserved servers in the data center and the multiple frequencies of the servers are dynamically set.Switch the internal state of the data center(server-off,multi-level sleep state,server idle state,and multi-level active state)for redistribution.Using the cloud computing environment simulation resource scheduling tool Cloud Sim to perform experimental simulations,the proposed algorithm in this thesis is compared with the baseline algorithm.Through the results of the experimental simulation,it can be found that this method can achieve lower energy costs,fewer application rejections,more stable task execution,and higher resource utilization.Regarding the resource management which is sensitive to the green energy,the goal is to optimize the energy consumption of task execution and the energy cost of power supply under the constraints of reliability and energy consumption.First,make corresponding predictions for the temporal and spatial characteristics of renewable energy.Then,the management strategy of heterogeneous energy is carried out according to the information related to the dispatching energy and the real-time load of the data center.Through the improvement of Cloud Simpy and the experimental simulation,the joint resource scheduling framework(JRSF)proposed in this thesis and the corresponding baseline algorithm is realized.The results obtained through experimental simulation found that this method can well adapt to renewable energy and data center loads which are dynamically changing,resulting in lower energy consumption,lower energy costs,and more stable system execution.