Fresh-Fi: Enhancing Information Freshness in Commodity WiFi Systems via Customizing Lower Layers

Enhancing information freshness in wireless networks has gained significant attention in recent years. To optimize or analyze information freshness, which is often characterized by the age of information (AoI) metric, extensive theoretical studies have been conducted on various wireless networks. Early research has demonstrated the significance of last-come-first-served (LCFS) packet scheduling and controlled status sampling (i.e., packet generation) in improving information freshness. These mechanisms have been widely adopted in subsequent studies. However, the effective implementation of these mechanisms in commercial off-the-shelf (COTS) wireless devices has not been thoroughly investigated, which could limit the practical application of information freshness-oriented protocols in real-world systems. Our work aims to address the gap by exploring the effective implementation of the information freshness-oriented mechanisms mentioned above in COTS WiFi devices that use the Linux operating systems. Our attempts reveal that the physical layer queue of WiFi devices operates on a first-come-first-served (FCFS) basis, and the packet generation process cannot be precisely controlled by default. To overcome these challenges, we develop Fresh-Fi, an information freshness-oriented protocol stack that involves careful customizations to the Linux networking protocol stack. Fresh-Fi mainly incorporates a mac80211 subsystem-based LCFS queue and a real-time kernel-based cross-layer tunnel between the mac80211 subsystem and the application layer for triggered packet generation. Our experiments show that implementing Fresh-Fi can significantly improve AoI performance. Specifically, we observed that Fresh-Fi improved AoI performance by over 13 times when compared to a baseline design that relies on an LCFS queue implemented in the application layer in standard Linux.