Research On Survivable Routing And Spectrum Assignment In Inter-datacenter Optical Networks

With the rising of high-bandwidth demand services such as public media,social networks,and cloud computing,the amount of computing tasks in datacenters and the traffic in Internet has grown exponentially.The traffic in inter-datacenter network is highly dynamic and unpredictable,and therefore the huge bandwidth requirement is challenging.Elastic optical networks can transmit large-capacity and allocate spectrum flexibly,which has become an inevitable choice for the physical layer technology of datacenter interconnection.The routing,modulation level,and spectrum assignment(RMLSA)play a key role in improving resource utilization and other performance.Therefore,it has become one of the key technologies in the management and control of optical networks.More importantly,the inter-datacenter elastic optical networks(IDC-EONs)carry a large amount of traffic,and network failures may cause huge economic losses.Therefore,it is important to improve the utilization of network resources while ensuring the survivability of the network.In this paper,we focus on the survivable routing and spectrum assignment in IDC-EONs.The main research work and innovations are as follows.(1)For survivable anycast RMLSA,we propose a scheme named shared backup path protection algorithm based on the minimum spectrum occupation combination strategy(SBPP-MSOCT).The existing spectrum occupancy of the network is considered and the working path and protection path with the smallest sum of spectrum consumption are selected,which can provide 100%protection in the single-link or single-datacenter failure scenario.Simulation results in NSFNET topology show that,compared with the reference algorithms,SBPP-MSOCT can reduce the request blocking probability by 71%and 90%,and improve the spectrum resource utilization by 14.5%and 19.4%.Moreover,it can reduce about 4 and 7 frequency slots for each successfully served request.(2)To address the routing and spectrum allocation for deadline-driven manycast requests,a manycast survivable routing,modulation level and spectrum assignment based on deadline-driven queuing with capacity weighting(MA-SRMLSA-DQCW)scheme is proposed.This scheme could achieve joint optimization in both frequency-and time-domains while providing 100%protection against single-link failure.Moreover,MA-SRMLSA-DQCW could construct working routing by light-tree and schedule the requests before their deadlines according to their capacity and maximum waiting time.Simulation results demonstrate that compared to Cheapest Insertion Multicast Tree with Fragmentation based on Shared Information Flow Protection and Dedicated Path Protection Tree,MA-SRMLSA-DQCW achieves high efficiency in reducing the request blocking probability with the maximum reduction ratio of 22.5%and 64.8%and offers a substantial reduction of capacity blocking probability with the maximum reduction up to 24.9%and 65.5%,respectively.What's more,MA-SRMLSA-DQCW scheme provides the same survivability with less spectrum occupancy.(3)In order to address the problem of joint allocation of computing resources and spectrum in IDC-EONs,we propose survivable routing,modulation level,spectrum,and computing resource allocation algorithm(SMRLSCRA).SMRLSCRA solves the spectrum and computing resources assignment problem of computing tasks with different communication methods.Moreover,we propose three datacenter selection strategies,namely shortest path first strategy(SPF),computing resource first strategy(CRF)and random destination strategy(RD).Simulation results show that,compared with SRMMLSCRA based on SPF(SRMMLSCRA-SPF)and SRMMLSCRA based on RD(SRMMLSCRA-RD),SRMLCRA based on CRF(SRMMLSCRA-CRF)can serve the largest amount of protected computing tasks,and the maximum reduction of maximum index of used on all the links is by 17.2%and 11.3%and the maximum blocking rate of the protected computing tasks is reduced by 29.2%and 28.3%.Besides,compared with SRMMLSCRA-CRF and SRLMLSCRA-RD,SRMLSCRA-SPF consumes the least number of spectrum per unit of computation and the maximum reduction is 16.2%and 35%respectively.SRMLCRA-CRF is applicable to networks with large business calculations.SRMLSCRA-SPF is suitable for network scenarios where the amount of business calculation is small and communication resources are scarce.SRMLSCRA-RD is suitable for network scenarios where service delay is sensitive due to low computational complexity.