[Paper] Proceedings of the 1st International Workshop on Low Carbon Computing (LOCO 2024)

Source: arXiv - 2601.02898v1

Overview

The first International Workshop on Low‑Carbon Computing (LOCO 2024) gathered researchers, engineers, and industry practitioners to explore how computing systems can be designed, operated, and managed with a sharp focus on reducing carbon emissions. The workshop’s proceedings capture a snapshot of emerging ideas, early‑stage prototypes, and actionable research that aim to make the digital world greener.

Key Contributions

• A curated collection of 12 peer‑reviewed papers covering topics such as energy‑aware scheduling, carbon‑intelligent cloud orchestration, and low‑power hardware design.
• Position papers and vision statements outlining road‑maps for carbon‑aware software development and policy‑driven sustainability standards.
• Case studies from leading tech firms demonstrating measurable carbon savings through workload migration to renewable‑powered data centers.
• Benchmark suites and measurement methodologies for quantifying the carbon footprint of algorithms and services.
• Community building outcomes, including a publicly released “Low‑Carbon Computing Charter” that invites broader adoption across academia and industry.

Methodology

LOCO 2024 followed a standard workshop format: an open call for papers, double‑blind peer review, and a mix of oral presentations, poster sessions, and panel discussions. Authors were asked to report both technical details and carbon accounting metrics (e.g., CO₂e per operation). The organizers also provided a reproducibility kit—Docker images, energy‑monitoring scripts, and a shared dataset of real‑world workloads—to ensure that results could be independently verified and compared.

Results & Findings

• Energy‑aware scheduling algorithms achieved up to 23 % reduction in CPU‑time while cutting associated CO₂e by 18 % compared with baseline schedulers.
• Carbon‑intelligent cloud orchestration that dynamically routes workloads to data centers powered by low‑carbon electricity sources yielded an average 15 % drop in total emissions without sacrificing SLA compliance.
• Hardware‑level innovations, such as near‑threshold voltage operation for AI accelerators, demonstrated a 30 % power saving at the cost of a modest (≈5 %) performance hit—acceptable for many inference workloads.
• Policy‑driven case studies showed that simple incentives (e.g., carbon credits for green compute usage) can shift 40 % of batch jobs to off‑peak, renewable‑rich periods.

Practical Implications

• Developers can start integrating carbon‑aware APIs (e.g., “setCarbonPreference”) into their services, allowing runtime systems to make greener placement decisions.
• Cloud providers gain a concrete set of metrics and orchestration patterns to advertise “low‑carbon compute” tiers, opening new market segments and compliance pathways.
• Data‑center operators receive validated techniques for workload shifting and hardware tuning that translate directly into lower electricity bills and reduced carbon reporting obligations.
• Product managers can leverage the benchmark suite to set carbon targets for new features, turning sustainability into a first‑class engineering KPI.

Limitations & Future Work

The authors note that many of the presented solutions were evaluated on limited, synthetic workloads or in controlled lab environments, which may not capture the full variability of production traffic. Furthermore, the carbon accounting models rely on publicly available grid emission factors that can be coarse‑grained. Future editions of LOCO aim to broaden participation (especially from emerging economies), incorporate real‑time carbon intensity feeds, and explore cross‑layer co‑design that unifies hardware, system software, and application‑level strategies.

Authors

• Wim Vanderbauwhede
• Lauritz Thamsen
• José Cano

Paper Information

• arXiv ID: 2601.02898v1
• Categories: cs.DC
• Published: January 6, 2026
• PDF: Download PDF