A recent addition to the Linux kernel, version 6.13, is set to revolutionize data center efficiency and reduce energy costs. Researchers from the University of Waterloo in Canada have developed a code update that optimizes how data is delivered to applications based on network traffic conditions, resulting in significant energy savings.

The update, which consists of approximately 30 lines of code, was developed by Waterloo professor Martin Karsten and Joe Damato, a distinguished engineer at Fastly. It is based on research published in a 2023 paper, which investigated kernel versus user-level networking and found that a small change could increase application efficiency while cutting data center power usage.

The new code adds a NAPI configuration parameter, irq_suspend_timeout, which helps balance CPU usage and network processing efficiency when using IRQ deferral and napi busy poll. This allows the system to automatically switch between two modes of delivering data to an application – polling and interrupt-driven – depending on network traffic, to maximize efficiency.

In polling mode, the application requests data, processes it, and then requests more, in a continuous cycle. In interrupt-driven mode, the application sleeps, saving energy and resources, until network traffic for it arrives, then wakes up and processes it. According to Karsten, the new mechanism is particularly useful in data center scenarios where server machines run a small number of dedicated server applications, which can be connected to a set of transmission queues in the NIC (network interface card).

Karsten explained that when network traffic is heavy, it is most efficient to disable interrupts and run in polling mode. However, when network traffic is light, interrupt-driven processing works best. "An implementation using only polling would waste a lot of resources/energy during times of light traffic. An implementation using only interrupts becomes inefficient during times of heavy traffic. … So the biggest energy savings arise when comparing to a high-performance always-polling implementation during times of light traffic," he said.

The implementation of the new parameter, as described by Damato in the patch cover letter, is efficient because it avoids softIRQ execution interfering with application processing during busy periods. It can be used with blocking epoll_wait to conserve CPU cycles during idle periods. The effect of alternating between busy and idle periods is that performance (throughput and latency) is very close to full busy polling, while CPU utilization is lower and very close to interrupt mitigation.

Enabling the feature requires a small tweak to applications and the setting of a system configuration variable. While Karsten cannot yet quantify the energy benefits of the technique (the 30% saving cited is best case), he believes that the biggest energy savings arise when comparing to a high-performance always-polling implementation during times of light traffic.

The implications of this update are significant, particularly in the context of data center energy consumption. As data centers continue to grow in size and complexity, energy efficiency becomes an increasingly important consideration. This update has the potential to make a substantial impact on data center energy costs, and its adoption could have far-reaching consequences for the industry.

As the tech industry continues to evolve and grow, innovations like this Linux kernel update will play a crucial role in shaping the future of data center operations. With its potential to reduce energy costs and increase efficiency, this update is an important step forward in the quest for more sustainable and efficient data center management.