 Abstract：In the computer networks,because of huge scale,uncooperative,heterogeneous and distributed administration,it is difficult to measure dynamic states and performance of the network directly.Therefore,it is of practical importance to develop a statistical way to infer some characteristics that cannot be measured directly.Network tomography has been regarded as one of the most promising methodologies for performance measurement and inference of the computer network.As a result of combination of communication network,tomography technology,and statistics,network tomography is a novel technology about measurement and inference of network performance.It can be used to infer unobservable network behaviors from endpoint measurement and does not require cooperation between the network internal elements and the end users.A quick overview of the conception and mathematical model of network tomography is presented.The uptodate research works on measurement and statistical inference technologies are introduced and some valuable network tomography technologies are analyzed explicitly.Finally,the research directions and open problems in this area in the future are provided. |
