Research On Bandwidth Scheduling Based On Global Optimization In High Performance Network

With the extensive informatization of science,engineering,and commercial applications,the scale of generated data has continued to increase,which has led to problems such as large-scale data transmission and its transmission service quality.Next-generation research networks such as Internet2 and Energy Science Network(ESnet)are being developed to provide high-speed on-demand data access between partner organizations.These high-performance networks typically employ bandwidth reservation services to provision dedicated high-bandwidth links to carry out predictable and reliable data transmission with end-to-end QoS assurance.The superiority of bandwidth reservation algorithms in high-performance networks directly affects network performance.This paper studies the following three aspects of bandwidth reservation in high-performance networks:(1)The problem of maximizing the total bandwidth(k> 1)of disjoint paths of k links is studied,and it is proved that this problem is a NP complete problem.A heuristic maximum K-path bandwidth algorithm(MKPBA)is proposed and implemented according to the idea of minimizing single resources in the shortest-widest path algorithm.MKPBA is implemented and compared with the modified version of the existing solution algorithm MBA(ImpMBA)to evaluate its performance.The simulation results show that the algorithm is superior to the comparison algorithm in terms of bandwidth value and the ratio of completing user requests.(2)Combining the advantages of both the fast scheduling response and the overall performance of periodic scheduling,the problem of maximizing the number of periodic scheduling in real-time scheduling is studied.A heuristic minimum hop count given bandwidth path scheduling algorithm(FBMHA)is proposed and implemented,and a greedy algorithm that uses the widest path algorithm scheduling,Greed-MSR,is proposed.A lot of experiments have been carried out on the FBMHA algorithm and the Greed-MSR algorithm.The experimental results show that the FBMHA algorithm has a significant improvement in the number of scheduling successes and the amount of data transmitted compared to the Greed-MSR algorithm.(3)A problem of maximizing the number of requests to be scheduled in periodic scheduling in high-performance networks is also studied.This problem is shown to be NPcomplete and difficult to approximate.A heuristic multi-path based flexible scheduling algorithm(FSABM)is designed and implemented.For performance evaluation,FSABM is compared with the OSCARS-MaxR-DC algorithm used by OSCARS and the best FBRmaxR algorithm for single-path scheduling.Experimental results show that FSABM achieves better performance than the algorithms in comparison in terms of overall scheduling performance,including success rate and the amount of transmitted data.