Sampling Rate Control with AGE-of-Information in Software-Defined Network Measurement

This paper considers a network measurement scenario where a controller requests updates (results of network measurement) from a switch. The updates are produced at the switch when a series of operations are finished such as collecting data packets and performing computations on them. The switch needs to send updates to the controller after receiving requests. The goal is to find an optimal time interval between two adjacent update requests of the controller, which is important as it can affect the accuracy and timeliness of network measurement results. Specifically, If the time interval is too short, the amount of information collected by the switch is insufficient, which affects the accuracy of the measurement results. On the other hand, it will cause the uploaded measurement results to be delayed and unable to reflect the current network status if the time interval is too long. Therefore, combining the theoretical framework of AOI to adjust the time interval between two adjacent update requests of the controller can optimize network status and is helpful for network management and control. First, this paper use an indicator called AOI (Age of Information) to model the timeliness and freshness of an update and model the quality of an update as a non-decreasing function of the time interval. Second, a convex optimization model is established by integrating AOI, quality and cost. Finally, it solve the convex optimization problem using the Lagrange Multiplier Method. We conduct extensive experiments and the experimental results show that the Average Relative Error and Sum AOI of our solution is obviously lower than baseline solutions. Furthermore, the Information Entropy of our solution is 23.9 on average while the other two baseline solutions are 22.5 and 7.4, which demonstrates the superiority of our solution.