Research On QoS Strategy And Implementation Method Of Web Server Based On Request Content

With the emergence of new technologies and new scenarios,network traffic has increased in scale.Large traffic impacts impose stricter requirements on server performance.It is particularly important to design and implement effective strategies to resist traffic impacts,and to provide different QoS(Quality of Service)for differentiated traffic.The improvement of service quality can be considered from two aspects.On the one hand,it is considered from the perspective of data forwarding: first classify the traffic according to the actual business scenario,and then forward the classified traffic according to the optimization strategy designed on the router side.On the other hand,from the perspective of the server: improve the service quality by increasing the server bandwidth and expanding the cluster.Although the above two aspects can alleviate the impact of traffic and provide a relatively reliable quality of service,it cannot fundamentally solve the problem.The forwarding optimization from network nodes is easily subject to the upgrade and maintenance of a large number of nodes,and the expansion of the server cluster will also be Due to capital,technology and other issues are restricted.Therefore,research on server-side QoS strategies can be considered,so that the server itself can use these optimization strategies to improve service quality within limited resources.The premise of QoS optimization is to classify traffic and then optimize QoS for different types of traffic.Traditional network traffic classification technology has been relatively mature,mainly divided into port-based traffic classification and deep packet inspection-based traffic classification.This thesis mainly focuses on the research of QoS strategy after classified traffic,and adopts the traffic classification method based on the content of the request.This strategy can support multiple traffic types.This thesis focuses on the classification and identification of more general resource types,then conduct follow-up QoS strategy analysis and research.Through the above discussion,on the premise of dividing the traffic type based on the content of the request,this thesis starts from the two technical indicators of service quality,delay and bandwidth,and analyzes and researches the server-side QoS related strategies to design the Web server delay strategy and bandwidth strategy.Realize,and then design the experimental platform by simulating the server process,through the verification experiment to verify and analyze the delay strategy,and the comparative experiment to compare the dynamic bandwidth allocation strategy based on reserved bandwidth with the traditional fixed bandwidth allocation strategy to analyze the strategy effect.The work of this thesis mainly includes the following points:(1)HTTP request classification and identification method design.The HTTP traffic requests are divided into different priorities based on the resource type and size.Then store it in different scheduling queues and prepare for the subsequent QoS strategy research.(2)Delay control strategy design.Adjust the delay time by changing the number of threads,dynamically control the delay time ratio of different priority queues,ensure that the system can reasonably allocate thread resources according to the delay ratio given by the configuration file for different request types,and reach the set target within the error range ratio.(3)Design of dynamic bandwidth allocation strategy based on reserved bandwidth.This theis proposes to set a shared reserved bandwidth for the request queue after classification,and dynamically control the bandwidth allocation by setting a bandwidth adjustment threshold.This strategy guarantees that different request types will not be "starved to death" due to competition.Secondly,the rational utilization of bandwidth resources avoids the waste of bandwidth and improves the overall quality of service.(4)Design of monitoring experiment platform based on Web Socket.The server data is pushed through Web Socket technology,and the performance indicators that need data collection are determined,the design of the entire monitoring platform is completed,and the operation effect of the delay strategy and bandwidth strategy is analyzed through the monitoring data during the experiment.The contribution of this thesis is mainly reflected in the aspect of bandwidth regulation.The traditional token bucket algorithm quantifies bandwidth resources through the concept of tokens.On this basis,this thesis proposes a dynamic bandwidth allocation strategy based on reserved bandwidth.This strategy sets the shared reserved bandwidth for different types of requests and the threshold for bandwidth adjustment.When bandwidth usage falls below the lower threshold multiple times,excess bandwidth can be stored for use by other types of requests.When the bandwidth usage rate exceeds the upper threshold multiple times,part of the bandwidth usage can be obtained from the reserved bandwidth.At the same time,a minimum guaranteed bandwidth is set for different types of requests to avoiding starvation of requests when dynamically adjust the bandwidth.Then,compared with the fixed bandwidth control strategy through experiments,the results show that the strategy can effectively improve bandwidth usage,avoid bandwidth waste,and improve server service quality.