Here’s the translation of your text into American English:
—
Researchers at the University of Waterloo optimize 30 lines of code to achieve more efficient CPU usage and significantly reduce energy footprint
A simple modification of just 30 lines of code in the Linux kernel could reduce energy consumption in data centers by 30% to 45%, according to a study conducted by the Cheriton School of Computer Science at the University of Waterloo (Canada). The research was presented at the ACM Sigmetrics 2024 conference and proposes a non-intrusive optimization that enhances network performance without the need to add new functionality to the operating system.
Optimization based on efficient cache use and interruption suspension
Professor Martin Karsten, one of the leaders of the study, explained that the adjustment is based on reorganizing the order of operations within the Linux network stack. This improvement allows for a better utilization of CPU caches, similar to rearranging processes on an assembly line to avoid unnecessary movements.
“We haven’t added anything; we’ve just changed the order of execution. It’s like redesigning an industrial plant to make everything flow more logically,” Karsten explained. The change leverages the IRQ interruption suspension mechanism, which allows for a reduction in the number of processor interruptions during peak traffic, thereby improving energy efficiency without penalizing latency during low-activity times.
An adjustment with global potential
The modification was developed alongside Joe Damato, a distinguished engineer at Fastly, and could have a significant impact on large data centers if adopted widely. Since Linux is the predominant operating system in server environments, its optimization could lead to energy savings on a global scale.
“If companies like Amazon, Google, IBM, Oracle, or Meta activated this feature, we could save gigawatt-hours of energy worldwide,” Karsten stated. “Every request that circulates on the internet could benefit from this improvement.”
The implementation is part of a broader project to create an eco-friendly server room in the university’s new mathematics building.
The open-source community supports the proposal
The adjustment has received positive feedback from various sectors of the open-source community. Ann Schlemmer, CEO of the open-source database company Percona, regarded it as an example of the collaborative potential of open-source software.
Also, Jason Soroko, an expert from the cybersecurity firm Sectigo, praised the proposal as “legitimate and having a high long-term impact,” highlighting that similar optimizations could be applied in other areas of the operating system.
Large-scale savings with small changes
While an individual network card doesn’t consume much energy, the effects are amplified on a large scale. According to Jamie Boote, a security consultant at Black Duck Software, a single server rack can consume around 160 watts just in network interfaces. Reducing this consumption across hundreds or thousands of racks is comparable to replacing an entire installation of incandescent bulbs with LED lights.
“The current Linux network stack was designed for much less demanding hardware,” Boote explained. “Reprioritizing system tasks based on traffic helps use energy more logically in modern architectures.”
Not suitable for all environments
Despite its benefits, the optimization is not universal, warns Ariadne Conill, a developer of Alpine Linux and co-founder of Edera. Its implementation requires manual adjustments using tools like ethtool, and is especially aimed at large-scale data center operators (hyperscalers) with bandwidth-intensive applications.
Furthermore, this approach may impact network latency predictability and alter resource allocation. Therefore, it is recommended as a configurable option, not as a default setting.
“It’s an interesting improvement, but it should be used in controlled environments where trade-offs can be effectively managed,” concluded Conill.
In summary
The Canadian research demonstrates that energy efficiency does not always require large investments, but can be achieved through precise code adjustments. In a world where computing accounts for about 5% of global daily energy, optimizations like this can make a significant difference.
The proposal also reignites the debate about efficiency in modern programming, where the focus has shifted from optimization to functionality. For Karsten and his team, it is time to return to that precision approach: “Sustainability must become a priority in software development,” he concluded.
Source: Technewsworld