Font Size: a A A

Research On Key Technologies And Methods Of Scheduling In Hybrid Container Cloud

Posted on:2019-12-20Degree:MasterType:Thesis
Country:ChinaCandidate:Q WangFull Text:PDF
GTID:2428330626450122Subject:Computer Science and Technology
Abstract/Summary:PDF Full Text Request
Container virtualization technology surpasses virtual machine virtualization technology due to its high efficiency and speed,and it also promotes the development of container clouds.The enterprise private container cloud is the enterprise's core infrastructure.It not only undertakes the enterprise core information services,but also optimizes resource allocation,reduces operating costs,and improves service performance.However,the resource supply capacity of private container cloud is limited,and it is unable to cope with the consequent surge in resource demand when corporate business surges.Therefore,the private container cloud needs to be expanded to a hybrid container cloud.The cooperation model of private container cloud plus public container clouds can solve the resource shortage problem for enterprise.The related technologies and methods of cloud-scheduling in hybrid container cloud are the key to implementing this cooperation model.The cooperation model of hybrid container cloud based on multiple public container clouds still needs further improvement.The private container cloud manages multiple access points of public container clouds in a centralized manner,which increases the burden on the enterprise and isn't flexible.Task scheduling and resource allocation mechanism on multiple container clouds cannot be completely in accordance with the mechanism within a single container cloud,and must comply with the corresponding trading environment.Performance bottlenecks and single point of failures are common problems in the supply of core services by a single entity: the trust mechanism based on trust center in public container cloud is not so credible;the master-slave network of enterprise's private container cloud cannot flexibly adapt to the scale of resource elasticity.The private container cloud still has room for improvement in terms of automatic deployment of physical nodes,resource reservation mechanisms,and task scheduling.The private container cloud under the environment of hybrid container cloud that we considered contains not only the CaaS(container-as-a-service)layer,but also the service proxy(DHTClient and DHTNode in the paper).The DHT(Distributed Hash Table)P2P network manages multiple service interfaces of public container cloud.A trusted mechanism based on public blockchain is adopted,and a trusted trading process based on the mechanism is designed.DHTClient as a proxy in the private container cloud selects resources from public container clouds,based on a mechanism of continuous reverse auction.We adopt a super-node-based P2 P network to manage physical resources of CaaS layer in the private container cloud,enhancing capabilities of resource management.We adopt automatic deployment and resource reservation mechanisms to control elastic scaling of private container cloud.The improved GA on super-node is for task scheduling and resource allocation,cooperating with other super-nodes when necessary.The optimization objectives are demand satisfaction,balance between supply and demand(resource utilization),and load balancing.Finally,we design some experiments and analysis to verify the feasibility of our research.In conclusion,managing multiple public container clouds based on DHT P2 P networks is more flexible.The smart contract and blockchain based trading mechanism will lead to higher trust and platform stability.The public container cloud selection mechanism based on continuous reverse auction allows the enterprise to overcome resource shortage and to own trading advantage.The super-node-based resource management and resource reservation mechanism in private container cloud has stronger resource management capability and elastic scalability.Task scheduling based on our improved GA can increase efficiency,stability,and cost-reducing.
Keywords/Search Tags:cloud computing, hybrid container cloud, P2P, continuous reverse auction, elastic scaling
PDF Full Text Request
Related items