Rule Caching in SDN-Enabled Base Stations Supporting Massive IoT Devices With Bursty Traffic

Software-defined networking (SDN) is recognized as a promising solution for the efficient management of numerous devices in the Internet of Things (IoT). To appropriately forward the incoming packets, SDN-enabled devices request the controller for traffic rules resulting in a significant delay. To avoid frequent communication with the controller and reducing the delay, these devices cache the rules as match action pairs in flow tables for a certain amount of time. This is referred to as rule caching. Flow tables are made of ternary content-addressable memories (TCAMs) capable of high-speed parallel lookup. Nonetheless, due to the high expenses and power consumption of TCAMs, flow tables have limited capacity and cannot store rules of all the users. Therefore, the assignment of limited flow tables is a challenge in IoT networks with a large number of users. In this article, we consider an SDN-enabled base station serving a set of users in a cell and is equipped with a finite-capacity flow table. We assume users' traffic obeys a bursty ON-OFF model and formulate the efficient allocation of flow table entries to users over time. It results in a mixed-integer nonlinear program. We introduce a change of variables that turns the problem into an integer linear program. This is still intractable and cannot support problems with large dimensions such as massive IoT networks. To handle its complexity, we design a time-efficient procedure with close-to-optimal performance. Finally, numerical results demonstrate the significant improvement achieved by the proposed scheme.