Optimized Design of Data Center Service Scheduling Method Based on ICN
Virtualization technology enables flexible deployment of data center services at any location,introducing challenges in effi-ciently scheduling service requests to these potential locations.This paper proposes a data center service scheduling method based on Information Centric Networking(ICN)called ICSS(Service Scheduling Based on ICN).Leveraging the ICN's identifier-address sepa-ration,ICSS provides distributed service scheduling directly at the network layer's data plane,supporting line-speed forwarding to make real-time decisions on service request scheduling.This approach addresses the shortcomings of traditional scheduling methods re-garding path extension and transmission latency.ICSS considers service request size and virtual machine resource loads,employing the Particle Swarm Optimization and Genetic Algorithm(PSO-GA)for service scheduling to optimize service response time,enhance virtu-al machine load balancing and improving network scalability.To seamlessly integrate with existing data center IP services,this paper designs a method for the conversion between ICN and IP network protocols,ensuring transparency for ICSS to clients and servers.Ex-periments demonstrate that compared to the traditional Weighted Least Connection Algorithm(WLC),ICSS reduces the average service completion time by approximately 21％,and variances in CPU and memory resource utilization are reduced by an average of approxi-mately 10％and 10％.Compared to the network-layer distributed service scheduling method CArDS(Compute-Aware Distributed Scheduling),ICSS exhibits lower service completion times,with average reductions of approximately 32％and 29％in CPU and mem-ory resource utilization variances,respectively.Additionally,ICSS achieves an average reduction of approximately 97％in forwarding table entries and an average reduction of approximately 98％in control overhead.ICSS demonstrates significant advantages in service completion time,load balancing,and scalability.
information-centric networkingdata center networkservice schedulingload balancing methodsprotocol conversion
万方数据
维普
