To satisfy various network requirements in the Internet of Things (IoT), software-defined networking (SDN) is viewed as an indispensable technology. Given the flow table resources in switches are limited, we proposed the periodic subflow-based proactive flow installation mechanism (PSPFIM) in our previous work. By separating subflows, PSPFIM can accurately detect the transmission periods and preinstall flow entry into switches before the transmissions arrive to reduce flow table occupancy. PSPFIM, however, is dysfunctional if transmission periods are at the millisecond level or subflows cannot be separated. In this paper, we propose the dynamic adjustment for PFIM (DAPFIM). DAPFIM is more widely applicable because the enhanced detection algorithm is less affected by noises and can detect transmission periods at the millisecond level. Besides, the adjustment algorithm can correct the error between an expected transmission period and a real transmission period. Through simulations, DAPFIM can detect transmission periods that cannot be detected by PSPFIM and other related work. Besides, DAPFIM averagely reduces up to 78% flow entry duration compared to other related work by changing the unit of the idle timeout from seconds to milliseconds, which saves much more flow table resources in switches.

为了满足物联网（IoT）中的各种网络要求，软件定义网络（SDN）被视为必不可少的技术。鉴于交换机中的流表资源有限，我们在先前的工作中提出了基于周期性子流的主动流安装机制（PSPFIM）。通过分离子流，PSPFIM可以准确地检测传输周期，并在传输到达之前将流条目预安装到交换机中，以减少流表占用率。但是，如果传输周期处于毫秒级别或子流无法分离，则PSPFIM将无法正常工作。在本文中，我们提出了PFIM（DAPFIM）的动态调整。DAPFIM的应用更为广泛，因为增强的检测算法受噪声的影响较小，并且可以检测到毫秒级的传输周期。除了，调整算法可以校正预期的传输周期和实际的传输周期之间的误差。通过仿真，DAPFIM可以检测PSPFIM和其他相关工作无法检测到的传输周期。此外，与其他相关工作相比，DAPFIM通过将空闲超时单位从秒更改为毫秒，平均可减少多达78％的流进入时间，从而节省了交换机中更多的流表资源。