Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

SCARIF: Towards Carbon Modeling of Cloud Servers with Accelerators

Shixin Ji, Zhuoping Yang, Xingzhen Chen, Stephen Cahoon, Jingtong Hu, Yiyu Shi, Alex K. Jones, Peipei Zhou

TL;DR

SCARIF presents an end-to-end, open-source framework for estimating embodied carbon in data-center servers with accelerators by reconciling vendor reports with a GDP-like linear model trained on high-level server configurations. It addresses gaps in existing tools that overlook peripheral components and accelerator-related emissions by validating across multiple vendors and proposing a scalable accelerator carbon estimator via a ratio-based framework with ACT baselines. The work demonstrates how design choices in server configurations and accelerator deployments impact total carbon, revealing scenarios where upgrade decisions can shift carbon footprints and challenging conventional estimates that favor rapid device refreshes. Overall, SCARIF enables more accurate, system-level carbon analysis to guide data-center design and policy, with practical implications for hyperscalers and enterprise deployments.

Abstract

Embodied carbon has been widely reported as a significant component in the full system lifecycle of various computing systems' green house gas emissions. Many efforts have been undertaken to quantify the elements that comprise this embodied carbon, from tools that evaluate semiconductor manufacturing to those that can quantify different elements of the computing system from commercial and academic sources. However, these tools cannot easily reproduce results reported by server vendors' product carbon reports and the accuracy can vary substantially due to various assumptions. Furthermore, attempts to determine green house gas contributions using bottom-up methodologies often do not agree with system-level studies and are hard to rectify. Nonetheless, given there is a need to consider all contributions to green house gas emissions in datacenters, we propose SCARIF, the Server Carbon including Accelerator Reporter with Intelligence-based Formulation tool. SCARIF has three main contributions: (1) We first collect reported carbon cost data from server vendors and design statistic models to predict the embodied carbon cost so that users can get the embodied carbon cost for their server configurations. (2) We provide embodied carbon cost if users configure servers with accelerators including GPUs, and FPGAs. (3) By using case studies, we show that certain design choices of data center management might flip by the insight and observation from using SCARIF. Thus, SCARIF provides an opportunity for large-scale datacenter and hyperscaler design. We release SCARIF as an open-source tool at https://github.com/arc-research-lab/SCARIF.

SCARIF: Towards Carbon Modeling of Cloud Servers with Accelerators

TL;DR

SCARIF presents an end-to-end, open-source framework for estimating embodied carbon in data-center servers with accelerators by reconciling vendor reports with a GDP-like linear model trained on high-level server configurations. It addresses gaps in existing tools that overlook peripheral components and accelerator-related emissions by validating across multiple vendors and proposing a scalable accelerator carbon estimator via a ratio-based framework with ACT baselines. The work demonstrates how design choices in server configurations and accelerator deployments impact total carbon, revealing scenarios where upgrade decisions can shift carbon footprints and challenging conventional estimates that favor rapid device refreshes. Overall, SCARIF enables more accurate, system-level carbon analysis to guide data-center design and policy, with practical implications for hyperscalers and enterprise deployments.

Abstract

Embodied carbon has been widely reported as a significant component in the full system lifecycle of various computing systems' green house gas emissions. Many efforts have been undertaken to quantify the elements that comprise this embodied carbon, from tools that evaluate semiconductor manufacturing to those that can quantify different elements of the computing system from commercial and academic sources. However, these tools cannot easily reproduce results reported by server vendors' product carbon reports and the accuracy can vary substantially due to various assumptions. Furthermore, attempts to determine green house gas contributions using bottom-up methodologies often do not agree with system-level studies and are hard to rectify. Nonetheless, given there is a need to consider all contributions to green house gas emissions in datacenters, we propose SCARIF, the Server Carbon including Accelerator Reporter with Intelligence-based Formulation tool. SCARIF has three main contributions: (1) We first collect reported carbon cost data from server vendors and design statistic models to predict the embodied carbon cost so that users can get the embodied carbon cost for their server configurations. (2) We provide embodied carbon cost if users configure servers with accelerators including GPUs, and FPGAs. (3) By using case studies, we show that certain design choices of data center management might flip by the insight and observation from using SCARIF. Thus, SCARIF provides an opportunity for large-scale datacenter and hyperscaler design. We release SCARIF as an open-source tool at https://github.com/arc-research-lab/SCARIF.
Paper Structure (14 sections, 4 equations, 6 figures, 3 tables)

This paper contains 14 sections, 4 equations, 6 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. Observations
  4. Vendor Reports Analysis
  5. Comparsion between ACT and PAIA tool
  6. SCARIF Modeling and Framework
  7. SCARIF Tool Overview
  8. Modeling Validation Across Vendors
  9. Embodied Carbon Cost Modeling for Accelerators
  10. Carbon Analysis From SCARIF
  11. Analysis and Experimental Setup
  12. Tradeoff in Upgrading Servers
  13. Carbon Cost for Systems with GPU and FPGA
  14. Conclusion and Acknowledge

Figures (6)

  • Figure 1: The comparison of embodied carbon estimation among ACT, SCARIF (ours), and the vendor reports on different server settings. The carbon costs are normalized to the vendor reports. SCARIF (ours) significantly improves the accuracy.
  • Figure 2: The comparison between reports of ACT and PAIA for the same Dell R740 server setup. Carbon cost in the figure are using kg as unit.
  • Figure 3: The overview of SCARIF tool.
  • Figure 4: Comparison between reported embodied carbon emission and predicted embodied carbon emission on 37 servers reported by Dell.
  • Figure 5: Carbon savings of upgrading vs. non-upgrading servers in four different states in United States.
  • ...and 1 more figures