| The emerging industries,such as Internet of Things,Unmanned Driver,Virtual Reality and so on,have put forward new demands in bandwidth,delay and reliability.The service mode of traditional cloud computing which transmits data to the core data center can not meet these new demands.Driven by the new requirements,edge computing with virtue of its proximity to users and optical network with high-capacity and high-bandwidth play an important role in achieving accurate transmission,real-time transmission and efficient processing of massive data.The storage and processing capacity of edge micro data centers deployed in edge networks is limited due to constraint of cost.Therefore,the collaboration of multiple edge micro data centers connected with optical network is the main form of edge computing.In the cooperative mode of edge micro data centers,a job request will contain multiple related parallel subjobs.The influence between multiple subjobs should also be considered when solving the resource allocation problem of concurrent jobs.Therefore,the distributed requests mode of"job-subjob" brings new challenges to the allocation of computing resources and network resources:on the one hand,subjobs need to be processed in local micro data centers in order to meet the bandwidth and delay requirements.Hence,how to determine the priority of multiple subjobs in a micro data center is a key issue.On the other hand,the communication in a core data center has becomes the communication between subjobs in multiple micro data centers,which increases the pressure of optical network between micro data centers.Hence,the effective allocation of optical network resources between micro data centers is another challenge.To solve these challenges,in the situation of edge micro data centers interconnected with optical network,this paper focuses on the joint optimization of jobs scheduling in a micro data center and lightpath provisioning between micro data centers for requests with deadline constraints based on the distributed requests characteristics of"job-subjob".The main work and innovations are as follows:(1)At the job level of distributed requests,a joint optimization algorithm of jobs scheduling and lightpath provisioning based on requests with deadline constraints is proposed.In order to maximize throughput,this paper has carried out the research works of jobs scheduling and routing,spectrum and modulation allocation for optical network,and proposes a joint jobs scheduling and lightpath provisioning algorithm.The urgency scheduling strategy is designed for the characteristics of requests with deadline constraints,and the delayed transmission lightpath provisioning strategy is proposed according to the "job-subjob" characteristics of distributed requests.The simulation results show that the request blocking rate is optimized by joint considering job scheduling and lightpath provisioning.The delayed transmission strategy can improve performance in terms of the blocking rate and network utilization,which saves the spectrum resources in optical nework interconnecting micro data centers.(2)At the subjob level of distributed requests,a jobs combined strategy and distance-path adapitive modulations scheme based interaction between jobs are proposed.Aiming at the resource optimization for jobs geographically distributed,a mathematical model considering deadline constraints and spectrum resource constraints is established,and a combined jobs scheduling strategy considering the interaction between jobs is proposed.Based on the combined jobs,a lightpath provisioning strategy with distance-path adaptive modulations is designed according to the overlap of paths.The simulation results show that the combined jobs scheduling and lightpath provisioning algorithm achieves a better optimization on blocking rate optimization,and can approach the optimal solution within more effective time complexity.Moreover,the proposed distance-path adaptive modulations can achieve the better tradeoff between blocking rate and cost consumption. |
PDF Full Text Request