Research On Key Technologies Of Adaptive And High Concurrency Network Topology Measurement

With the rapid development of the Internet,the scale of the Internet has become larger and larger,and its structure has become more and more complex.Exploring the network topology is of great significance.Network topology measurement is the basis for research on optimization of network resource allocation,network protocol design,network attack detection and prevention,network geolocation,etc.It is an important prerequisite for network space surveying and mapping and situational awareness.This article will study the adaptive and high-concurrency network topology measurement technology.The specific research content is as follows:(1)In order to pursue the goal of less measurement packets,an adaptive network topology measurement method is proposed.Aiming at the problem of the low number of packet responses for a single packet type detection technology,a multi-type packet detection method(MTPP,Multi-Type Packet Prober)is proposed,that is,the type of the detection packet is dynamically modified according to historical detection data.In order to solve the problem of packet redundancy in subnet discovery,a Tree Subnet Found(TSF)method is proposed,that is,the tree branch is continuously performed during the detection process to reduce detection redundancy.Further,in order to reduce the problem that the detection of a single target is not applied to historical detection data,a combination of the subnet position reference method and the forward and reverse detection method is proposed,that is,the similarity of the neighboring positions of the subnet is used to select a similar path as a reference for forward and reverse.Combination detection.Aiming at the packet redundancy problem in subnet discovery and single target IP address multi-task measurement,a global reference set reference method is proposed,that is,recording historical task detection results,which are used as a reference for current task detection to reduce redundancy.Develop self-adaptive traceroute(SAT,Self Adapt Traceroute),apply global reference set reference method,reduce the redundancy of single target IP address multi-task network measurement packet.The experimental results show that MTPP is about 13.6%higher than the target response of single packet detection technology;the redundancy of TSF is R = 4.5%,and the number of probe packets sent is about 83.7% lower than the number of probe packets discovered by the Scamper subnet.Global reference is applied.After integrating the reference method,the number of detection IP addresses is reduced by about 96.1%,and the number of detection packets that need to be detected is reduced by about 96.0%;the redundancy of SAT is R = 0.69 <1,and the number of detection packets is reduced by about 32.2%.(2)In order to pursue the goal of high detection efficiency,a high-concurrency network topology measurement method is proposed.Implementation of highconcurrency network topology measurement technology faces many difficulties,such as the design and sharing of task status,the distinction of different granular storage structures,the decoupling between detection modules,the high-concurrency realization of single-process and single-thread,the design of different levels of concurrency,and multiple Differentiation of detection type levels and compatibility of adaptive methods,etc.In order to cope with the above difficulties,this article proposes a framework and four methods.One architecture refers to the highconcurrency hierarchical network topology measurement architecture;and the four methods refer to the storage granularity distinguishing method,the polling highconcurrency,high-concurrency tree branch subnet discovery method(HCTSF,High Concurrency Tree's Subnet Found),and the dispatch train(Packet Train)measurement method.Experimental results show that the detection efficiency of a single target IP address(Packet Train)is about 90.0% higher than that of traceroute,and the efficiency of High Concurrent Tree Branch Subnet Discovery(HCTSF)is about 18.5% higher than Fastrace.(3)A high-concurrency network topology measurement system with a single process and a single thread is realized,and the customization of detection methods is supported.The system supports customization of detection methods and supports five detection methods: MTPP,TSF,SAT,HCTSF and Packet Train.The MTPP experiment is designed.The experimental results show that MTPP is about 13.6% higher than the single packet detection technology target response;the test detection packet redundancy results show that the system supports both SAT(single target multi-task)and TSF(subnet discovery).It has low packet redundancy;the test performance results show that the detection efficiency of HCTSF(High Concurrent Subnet Discovery)supported by the system is increased by about 18.5%,and the detection efficiency of Packet Train(High Concurrent Single Target Multitask)is increased by about 90.0%.