Research On Reliable And Rapid-Response SFC Scheduling Mechanism In Edge Networks

Virtualization technology enables cloud applications to share resources ef-ficiently.Virtualization,softwareization,and "cloud" have become the devel-opment trend.Software-Defined Network(SDN)and Network Function Virtu-alization(NFV)give the traditional telecommunications network unprecedented flexibility.Simultaneously,connecting multiple Virtual Network Function(VNF)in series into a Service Function Chain(SFC)expands network functions to meet the increasingly personalized network requirements of applications.Edge com-puting extends the cloud to users.Close to the user,edge cloud reduces delay-sensitive services' response time and improves the Quality of Service(QoS).However,the edge network causes challenges to the service function chain scheduling.The edge network fragmentation causes the communication delay between the VNF instances of the service function chain and affects the service function chain's final response speed.A reasonable scheduling mechanism is necessary to keep the strengths and avoid weaknesses of edge computing.Re-liability is also a critical QoS indicator.Providing continuously and stably SFC services in the unstable edge cloud network is also a challenge.The SFC reliability guarantee and response time optimization have re-ceived extensive attention and obtain many valuable results.However,there is still improvement room for modeling,optimization mechanism,and algorithm for the edge network's response time and reliability.This thesis's contributions are as follow:1.This thesis proposes a delay-sensitive geo-aware service function chain scheduling algorithm to optimize SFC's response time in the edge cloud.Due to resource preference and request location restriction,the VNFs that make up the service function chain are dispersed into multiple geographically separated edge cloud regions.The edge cloud reduces the communication delay to the user's location,but the communication delay between separated edge clouds is considerable.The communication delay between geo-cross edge clouds is the main bottleneck in the service function chain's response time.This thesis proposes the geo-aware,delay-sensitive service function chain placement algo-rithm,which adopts a critical path replication strategy to eliminate the cross-edge cloud communication delay caused by the service function chain topol-ogy.Besides,this thesis proposes a geographical-aware delay cost model to evaluate the service function chain placement solution's response time.Based on the multi-agent system and improved contract network protocol,this thesis realizes the service function chain placement solution and optimize response time through mutual game and negotiation between agents.Compared with the classic service function chain placement mechanism,the proposed mechanism optimizes the response time of service function chains of different topologies by 3%to 20%.2.This thesis proposes a graph-based service function chain reliability model and placement optimization algorithm to guarantee the service function chain's reliability.The algorithm first proposes a reliability evaluation model that considers the reliability of software and hardware and the length of the communication link.The Markov communication distance model can quickly estimate the communication distance between nodes in the complex edge cloud network.The solution space of the service function chain placement algorithm is an uneven non-Euclidean space.Therefore,this thesis proposes an improved heuristic search algorithm and a speed update mechanism based on topological distance based on graph theory.The experimental comparison verifies that the proposed mechanism improves 75%convergence speed of the algorithm and 10%performance of the service function chain.3.This thesis proposes a joint optimization mechanism of reliability and response time for service function chain scheduling in an edge network.Re-liability and response time are both critical QoS indicators.However,in the edge network,the two indicators have the opposite preference for the service function chain's placement.Reliability guarantee algorithms tend to allocate service function chains on nodes closer to the core network with higher relia-bility.However,the closer to the user,the faster the response speed.So,the response speed optimization mechanism tends to be closer to the user's edge node.This thesis models service function chain placement in edge networks as a joint objective optimization problem under resource constraints.This thesis proposes a joint optimization strategy for response time and reliability based on critical VNF replication.Besides,based on the edge network's characteris-tics,this thesis improves the multi-agent system,takes a slight edge cloud as a whole for resource scheduling,and introduces a middleman agent to preprocess the service function chain topology and refine the topology.Based on the eval-uation model of response time and reliability optimization,the agents evaluate the optimization of the replication strategy for the service function chain's reli-ability and response speed.In the comparison experiments,the proposed joint optimization mechanism guaranteed 99.999875%reliability while also improv-ing the response speed by 10%.