Fine-Grained Data Flow Scheduling Algorithm In Optical Cut-through Switches

On the road to the development of optical communication,optical packet switching technology has the advantages of high bandwidth resource utilization,flexible switching transmission configuration in units of packets,and ability to provide end-to-end connectionless transmission.It is regarded by most researchers as the most advantageous.In addition,optical packet switching technology can transfer existing network services to the optical network for transmission,and adopts wavelength division multiplexing(WDM)technology to realize the rational use of network bandwidth resources.Therefore,the status of optical packet switching technology in all-optical networks cannot be underestimated.With the introduction of optical cut-through(OpCut)switches using electronic buffers,optical packets can reduce the optical-electrical-optical(O/E/O)conversion frequency through cut-through switches to achieve high energy efficiency.Based on OpCut switch and combined with WDM technology,this thesis proposes a fine-grained data flow scheduling algorithm,which can improve the packet cut-through ratio of the switch under high WDM density.Research on the packet scheduling algorithm of OpCut switches has been extensive,among which the scheduling algorithm of coarse-grained switch flow performs better.The scheduling algorithm maintains the order of packets,which proves that the performance of packet delay is close to the ideal WDM output queuing switch.However,because the packet scheduling algorithm runs on the coarse-grained switch flow,the traffic congestion problem becomes very serious under the high WDM density.Therefore,this thesis proposes a subflow partitioning idea for OpCut switches to alleviate the flow congestion problem.From maintaining the order of coarse-grained switch flows to maintaining the order of fine-grained subflows,the scheduling algorithm becomes more refined and also does not affect network application performance.For the packet scheduling of OpCut switches subflows,a scheduling optimization problem model is constructed.The goal is to maximize the transmission wavelength of the subflow,so that more packets are transmitted to the destination port,thereby increasing the packet cut-through ratio.And this process is defined as the maximal cut-through(MCT)problem,through the reduction of the maximum independent set problem,the problem is proved to be NP-hard.Then based on the analysis of the MCT problem,a heuristic scheduling algorithm was designed for the WDM OpCut switch:the cut-through first(CTF)algorithm,which allows the packets that can be cut-through to be transmitted preferentially.Finally,this thesis uses three traffic statistics models and captures the real network traffic trajectory to evaluate the CTF algorithm in simulation experiments.The results of a large number of simulation experiments in this thesis show that under the proposed fine-grained data flow scheduling,the WDM OpCut switch achieves more than 60%higher sustainable traffic load,and reduces energu per bit by more than 30%,which proves the effectiveness of the fine-grained data flow scheduling algorithm proposed in this thesis.It also makes WDM OpCut switch a promising candidate for future optical packet switching networks.