Services that require high real-time performance, such as virtual reality (VR), online games, remote control of drones, and self-driving cars, are becoming increasingly popular and network devices require low-latency data transfer. However, performance tuning and data-transfer methods that sacrifice power consumption are often adopted to achieve
low latency data transfer. Thus, reducing power consumption is an important issue.

This paper proposes a packet-forwarding method that simultaneously achieves low latency, high throughput, and low power consumption for the widely used Linux kernel on general-purpose servers. Specifically, the proposed method is unique in that to address the latency problem of conventional packet-receiving methods on the Linux kernel, the authors adopt a polling model that is expected to achieve low latency and high throughput, and to address the power consumption problem of the polling model, they introduce a method to control sleep and wake up polling threads using hardware interrupts. These features are implemented in a highly practical and operable manner while guaranteeing high performance in the highly constrained Linux kernel.

It has been confirmed that the proposed method reduces the maximum delay time from the order of milliseconds to the order of microseconds and achieves 1.4 to 3.1 times higher throughput than conventional methods, while achieving power saving. Since the proposed method can significantly reduce latency and improve throughput while achieving power saving, and since it can be easily used in the widely used Linux kernel, it is expected to be used in many applications, including the aforementioned services that require high real-time performance, and is expected to become a cornerstone for the further development of the information and communication industry. For these reasons, this paper is highly evaluated as an appropriate paper for this award.