With the proliferation of cloud services and development of fine-grained virtualization techniques, the Cloud Management System （CMS） is required to manage multiple resources efficiently for the large-scale, highden...With the proliferation of cloud services and development of fine-grained virtualization techniques, the Cloud Management System （CMS） is required to manage multiple resources efficiently for the large-scale, highdensity computing units. Specifically, providing guaranteed networking Service Level Agreement （SLA） has become a challenge. This paper proposes MN-SLA （Modular Networking SLA）, a framework to provide networking SLA and to enable its seamless integration with existing CMSes. Targeting at a modular, general, robust, and efficient design, MN-SLA abstracts general interacting Application Programming Interfaces （APIs） between CMS and SLA subsystem, and it is able to accomplish the integration with minor modifications to CMS. The evaluations based on large scale simulation show that the proposed networking SLA scheduling is promising in terms of resource utilization, being able to accommodate at least 1.4x the number of instances of its competitors.展开更多
文摘With the proliferation of cloud services and development of fine-grained virtualization techniques, the Cloud Management System （CMS） is required to manage multiple resources efficiently for the large-scale, highdensity computing units. Specifically, providing guaranteed networking Service Level Agreement （SLA） has become a challenge. This paper proposes MN-SLA （Modular Networking SLA）, a framework to provide networking SLA and to enable its seamless integration with existing CMSes. Targeting at a modular, general, robust, and efficient design, MN-SLA abstracts general interacting Application Programming Interfaces （APIs） between CMS and SLA subsystem, and it is able to accomplish the integration with minor modifications to CMS. The evaluations based on large scale simulation show that the proposed networking SLA scheduling is promising in terms of resource utilization, being able to accommodate at least 1.4x the number of instances of its competitors.
