Research On Virtual Resource Allocation And Management For Edge-Cloud Mixed Environment

Cloud computing aggregates all kinds of resources into a shared resource pool through virtualization technologies,which can realize universal,on-demand and convenient access to the shared pool of configurable computing resources(such as networks,servers,storage,applications and services).Edge cloud is a new paradigm that aims to push computing services and storage contents near users(for example,in base stations,access points or aggregation network),to reduce latency,improve the quality of experience,and ensure efficient network operation and service delivery.Although the edge computing has the great potential to relieve the burden on core networks,its main bottleneck is the limited computation and communication capacities as compared with the cloud computing.In order to provide better service quality,the integration of cloud computing and edge computing is the development trend of cloud computing.When deploying applications in edge-cloud mixed environment,large-scale,extremely complex and high-speed data bring new technical requirements,including data acquisition,data storage,data organization,data analysis and real-time data release.Obviously,due to the contradiction between resource limitation and the pursuit of service quality,resource allocation for applications in cloud-edge mixed environment has become an important challenge.Based on the background of the integration of cloud computing and edge computing,this paper studies the resource allocation problem for cloud-edge mixed environment.It mainly analyses the resource scaling-out problem in the stream big data analytics when the input data increases steeply,the service function chain(SFC)embedding problem in Internet of Things(IoT)based on network function virtualization(NFV),and the resource allocation in distributed NFV with mobile edge computing.Our work focuses on these problems and presents the following contributions:1.For stream big data analytics,a participated task always needs to scale out resources when its input data increases steeply.To solve it systematically,a non-cooperative game model is designed to improve the platform parallelism,which describes the competitive relationships between tasks through a weighted total cost model.Using this model,a guiding big data analysis framework is established to analyze the resource scaling-out problem.The topology structure and task features in different scenarios are considered in the model,including simple topologies and complex topologies.In this model,the sum of the costs of all tasks is defined as the social cost of resource scaling-out,and the concept of exit equilibrium is proposed to describe the equilibrium of individual participant exit.Then we introduce the concept of price of anarchy(POA)to this game and get its upper bounds under specific conditions to describe the efficiency loss of Nash equilibrium.In order to achieve social optimum rather than sub-optimum,two economic classic tax-based incentive mechanisms:Pivotal Mechanism and Externality Mechanism are applied,to stimulate the participation of tasks.We make simulations in different scenarios including node degree and different characteristics of tasks.The simulations results show that our resource scaling-out mechanism can achieve a better performance close to social optimality.2.A deep reinforcement learning(DRL)based SFC embedding scheme is proposed as a solution,which combines the latest advances in DRL to solve the SFC embedding problem in dynamic and complex IoT scenarios.Some complex virtual network functions are decomposed into a set of virtual network function components and internal connection graphs.Then the virtual network function forwarding graph is constructed in NFV-enabled IoT.The environment state,action space and reward function of SFC embedding problem are given in this model.In addition,experience replay and target network in DQL are used to improve the convergence performance of this scheme.Our simulations consider different types of substrate network topologies.The simulation results show that the proposed dynamic SFC embedding scheme outperforms the existing schemes.3.A blockchain-based consensus protocol for distributed NFV in edge cloud is proposed,with detailed consensus steps and theoretical analysis.The blockchain is acted as a trusted third party,which collects and synchronizes messages for resource management nodes in different network environments.In order to improve the efficiency of resource allocation,the throughput of blockchain system and the total cost of implementing services in distributed NFV are both considered in this work.The view change,access selection and resource allocation are formulated as a multi-objective optimization problem.We use Dueling DQL to solve this problem.Simulation results are presented to show the effectiveness and convergence performance of the proposed scheme in different cases.