Research And Implementation Of Network Performance Management System Based On The Dynamic Threshold

Network performance management confirms CPU load, network flow in system, transmission rate, bandwidth availability ratio, delay, packet loss of the devices in network through monitoring all of the network devices, and find the problems of physical connection and system configuration, then optimize network performance and improve network operation efficiency with technologies such as network state monitoring and route optimization. There are several parts in network performance management, collect current data of network devices and connections, analyze related data and distinguish use trend, set threshold of performance index, make the network achieve the best condition with network management tools.An important step of the network performance management is threshold control, we can set network thresholds for various index which influencing network performance, such as CPU load, network flow. Once the threshold is set, when some performance index is passing the threshold, the performance management tools will report to the management. Currently, there are plenty of such products in market can meet customer demand on performance monitoring and also do use the static threshold strategy or dynamic threshold strategy. However, the so called dynamic threshold are basically calculated by baseline, which means the system will make regular evaluation on the busy degree of monitoring resource then take the outcome as threshold standard for next period of time. Such kind of threshold seem is dynamic in long-term but actually is still static in certain period of time. This paper will take it as entry point, propose new dynamic threshold calculation method based on exception amendment then develop a relatively integral network performance management system.Because exception amendment based dynamic threshold will take the historical performance data as analytical basis, any exception exist will affect the accuracy, therefore an amendment action must be taken and replaced by correct data at that historical point. Exception amendment based dynamic threshold is constant changing, it’s not simple average about historical performance data, the final gained threshold curve is fluctuating rather than straight line. Meanwhile, in order to give a certain of prediction ability on performance trends, the liner regression has been used. By doing this, the threshold curve not only can be used as test standard whether performance index has exceed the maximum, but also as trends prediction.In order to fulfill the exception amendment based threshold strategy, this paper design a complete network management system, this system not only can perform the unified configuration management about underlying basic strategy, but also with an liberalized present layer to meet different demands of customers, then ensure the effectiveness and efficiency of network performance management.