Research On The Key Technology Of Re-configurable Network Measurement

It is widely accepted that network measurement technique is an indispensable part of current and future network research. It serves as the basic means of in-depth understanding of network outside behavior and inside characteristics. Meanwhile it is also effective for network monitoring, maintenance, optimization and control.In the recent twenty years much has been achieved in this research field.However, with new network characteristics represented by expending Internet scale, high-speed network rate and a great variety of network service, the deficiency in flexible architecture and efficient resource usage suggests itself. Accordingly, network measurement technique should be further explored to accord with the development of Internet.Under this research background, supported by the 973 project named "research on the Architecture of the reconfigurable Fundamental Information Communication Network", this dissertation attempts to find a new valid method to improve the flexibility and the efficiency of network measurement. To solve the essential question of constructing diverse measurement ability based on limited measurement resources, the paper proposes the re-configurable network measurement technique, and studies such key problems as the measurement model, measurement task deployment and measurement algorithm implementation.The purpose of this article is to supply a set of technical proposals to construct a flexible and efficient measurement system for the Internet.The re-configurable network measurement technique can construct various and flexible measurement functions based on the limited measurement components according to the whole network measurement service requirements by network measurement task optimal deployment. This new measurement technique solves the problem of supporting diverse measurement requirements on limited resource totally and makes great progress on the performance of constructing measurement ability dynamically, deploying measurement tasks concurrently and using measurement resource efficiently. The main work and achievements of this article are as follows:1.To improve the deficiency of flexible architecture and efficient resource usage in current network measurement, a novel Network Measurement Re-Configuration model (NMRC) is devised.Based on the ideal of reconfiguration, NMRC decouples the measurement logical measurement control from the measurement function units. The model uses the measurement component as the basic reconfigurable unit and realizes the reconfiguration of measurement ability through the method of multi-granular hierarchical reconfiguration and measurement component dynamic combination.NMRC is composed of three logical function layers including measurement device layer, measurement adapting layer and measurement application layer. The device layer abstracts the general resource for measurement components and generates the resource views. While the application layer converts the measurement applications to the exact measurement requirements. Then the adapting layer adapts the measurement requirements to the measurement components so as to construct flexible measurement function and performance on the limited resources.Simulated results show that the success ratio of NMRC is higher than non-configurable models on averages 49.7% under different number of deployed tasks. Meanwhile, the average utilization rate of resource on measurement nodes is above 90% at most and the balance coefficient remains below 0.1.2.In order to solve the problem of adapting diverse measurement requirements to limited measurement resource flexibly and efficiently, the paper proposes measurement task deployment model in the re-configurable measurement architecture and also presents a Measurement Task Deployment optimization algorithm based on the Preferred deployed network Mapping (MTDPM). The paper introduces a resource function fitting method to describe the measurement deployment process in detail the best measurement component combination function for each measurement task is determined based on network and resource status.Then the problem of measurement task deployment is converted to the network mapping problem between the preferred deployed network and the measurement network. Finally a heuristic approach is utilized to optimize the deployment problem according to the three deployment principles including balanced usage of resource, measurement cost minimization and prioritized links mapping. The experimental results indicate that our algorithm’s success ratio of deployment is no less than 92% even in the condition of large number of deployment tasks and high task confliction probability. Also the average delay time of task sees an obvious decrease.3.As to reconfiguring measurement function and performance on measurement nodes, the paper presents the Reconfigurable Sampling algorithm based on the Measurement Component Combination relationships (RSMCC).This algorithm designs six measurement components and defines different combination relationships. Beside how to adjust the parameter values and configure the combination relationship for different sampling function and performance is defined. Our algorithm not only reconfigures the sampling function and performance according to the requirement, but also ensures the sampling results can satisfy the different measurement requirements. Furthermore theoretical analysis of fairy sampling error is deduced and the upper bound of theoretical error is obtained.The experimental results also prove that the resource utilization fluctuates little while the measurement environment changes greatly. The resource utilization always keeps upon 95% in different composing situation of network flows.4. Considering that the openness of measurement components may influent the re-configurable measurement system’s correctness and reliability, a novel measurement component conformance testing method is proposed. The paper starts at designing a Measurement Component state Transfer Model(MCTM)to describe the components working stream transfer process. Based on the model, a conformance test consequence generation algorithm (CTBMCTM) is devised.The algorithm derives a capacity network from measurement component transfer model and then constructs an Euler graph based on the capacity network using maximum flow marked algorithm.The input and output sequences on the Euler path of the Euler graph is the final conformance testing sequences. The simulated results indicate that among test sequences generated by our algorithm under different component combination topologies and scales,the sequence length is almost same to T&GS and the sequence generation time is 84.6% and 96.01% less than T&GS.Meanwhile, the approach can discover the abnormity measurement component exactly.