Design And Implementation Of An Available Bandwidth Monitoring Model

With the non-stop development of computer science and network technology, cloud computing and other novel technologies have come into our sights. While providing various high-performance services, cloud computing technology asks more from the underlying transmission network and requires a network monitoring system with illustrious performances.Considered the requirements above, the thesis attains a wholesome comprehension of monitoring techniques while researching the subject of bandwidth, which is one of the most prevalent network performance metrics in cloud computing. This thesis focuses on the theories and methods of bandwidth measurement technology, the types and features of network monitoring and several forecasting analytic models of bandwidth. The research in this thesis includes the following aspects.First, the thesis analyzes the traditional ways to measure bandwidth. Bandwidth measurement technology based on the delays of probing packets was the very first relevant technology. But it ignores the burst of the background traffic. Algorithms based on that technology cannot give accurate results because the errors created by background traffic are not easy to eliminate. Statistics-based probing techniques remarkably improves the accuracy of measurement results for not demanding high speed access and apt not to be affected by background traffic.Then, most applications under the circumstance of cloud computing are QoS-sensitive and require high speed network. Though algorithms based on statistics are accurate, they are incapable of adjusting the probing process. Thus, when available bandwidth becomes inadequate, probing traffic would probably disturb cross traffic. Taken the above into consideration, the thesis presents an Adjusting End-to-end Available-bandwidth Measurement algorithm, which implements self-adjusting functions with alarm components.Thirdly, the thesis studies network monitoring and forecasting techniques. In order to monitor the real-time available bandwidth of a network and forecast the trend of the monitored metrics, this thesis presents a model to monitor and forecast network bandwidth. The model is divided into several modules in accordance with monitoring requirements. Measurement module measures the available bandwidth, analysis module processes the measurement results and feedback module decides the trend of available bandwidth with the processed results.The model is implemented in CAOS at last. All modules in the model are re-encapsulated and patched into the cloud resources monitoring model. The extended functions of the cloud resources monitoring model include available bandwidth measurement and forecast. To verify the efforts of the model, a couple of experiments are conducted in cloud servers to testify the measurement accuracy, self-adjusting function and forecasting performance of the model. The consequences are positive. The model does well in monitoring and forecasting available bandwidth.