Research And Development Of Network Quality Of Service Measurement Method For Data Center Coverage Area

Network Quality of Service Measurement in Data Center Coverage Area is an emerging direction in the field of Network Measurement,which has received very high attention in the industry,for example,Microsoft has proposed a solution based on passive measurement at SIGCOMM 2019,but the harsh conditions have led to the inability of other cloud service providers to replicate it.Therefore,the data collection method of this topic uses active measurement,but at the same time,it faces very prominent problems.One of them is the complexity of the network conditions in the coverage area and the low purity of the measurement data.The quality of the network from the data center to the coverage area is assessed not for a single IP but for a set of IPs.Therefore,if the IPs are not properly categorized,it is not only difficult to extract useful signals from a large amount of noise,but also confuse a large number of irrelevant signals,resulting in secondary pollution.Second,the active measurement adopted in this thesis will bring high measurement cost.The number of measured IPs in the coverage area is extremely large,ranging from millions to tens of millions or even hundreds of millions,so how to control the measurement cost within an acceptable range and at the same time ensure that the measurement results meet the expectations is a problem that cannot be bypassed.Aiming at the above problems,this thesis proposes a network quality of service measurement method for the data center coverage area,which divides the measurement into the path measurement from the data center to the network in the coverage area and the QoS measurement from the data center to the IP at the end of the coverage area,and uses the AS-Geo Path and the IP address block which is announced externally to classify the IPs in the coverage area.In addition,while controlling the measurement cost,it also takes into account the measurement effect,in order to provide effective and reasonable measurement data to the operation and maintenance personnel,so that they can evaluate and monitor the network quality of service of the data center to its coverage area based on the measurement data,and rely on the historical measurement data for traceability in case of degradation or failure of the network quality of service.Practical tests show that the key technical points in the proposed measurement method are feasible and effective.The main work of this thesis is divided into the following aspects:(1)For the path measurement from data center to coverage area network,firstly,this thesis proposes a method of IP address block division in the coverage area,which divides IPs with the same externally announced network segments into the same address block according to the full amount of path measurement data in the coverage area.Then,this thesis proposes a method of path sampling measurement from the cyberspace,i.e.,only part of IPs are extracted from each divided address block to be measured,and at the same time,the thesis proposes an index for evaluating the effectiveness of path sampling measurement.At the same time,the thesis proposes an index for evaluating the effectiveness of path sampling measurement,so that users can decide the configuration of path sampling measurement after making a trade-off between the measurement cost and the effectiveness of sampling measurement.(2)For the QoS measurement of IPs from the data center to the end of the coverage area,the differences in the ability to discover active IPs when probing with different network protocols and different destination port numbers in the coverage area are first analyzed to determine the measurement configurations that can discover the maximum number of active IP addresses;and then,in combination with the results of the segmented address blocks,each IP in the coverage area is categorized into the address block corresponding to it.The measurement cost when each address block meets the data sample size required for statistics(generally greater than 30)is analyzed,and the measurement algorithm after optimization is proposed,which achieves the effect of trying to meet the data sample size statistical requirements of most address blocks while reducing the measurement cost.(3)Based on the above proposed measurement method,this thesis designs and develops a measurement system applicable to the scenario of this topic.This includes the design and development of a configuration center that stores measurement configurations,a measurement application that performs measurement tasks,and a data analysis and storage application.In addition,the key technical points in the proposed measurement method are tested using actual measurement data,and the feasibility and effectiveness of the path measurement method and the end IP measurement method for the data center coverage area are verified.