Pith. sign in

REVIEW 1 major objections 7 minor 93 references

Energy-Aware Workflow Execution: An Overview of Techniques for Saving Energy and Emissions in Scientific Compute Clusters

T0 review · 1 major / 7 minor · reviewed 2026-08-07 · deepseek-v4-flash

Pith's one-line read Scientific workflows running on compute clusters carry a measurable carbon cost, and a wide set of techniques already exists to cut it — from energy-efficient code generation and hardware choice to consolidation and carbon-aware scheduling.

desk verdict A useful survey with a genuinely load-bearing but fragile 4.5x energy-efficiency comparison that needs error bars before it should be cited. read the letter →

arxiv 2506.04062 v1 pith:DQA6V2GE submitted 2025-06-04 cs.DC

classification cs.DC
keywords carbonfootprintestimationscientificworkflowsenergy-awareschedulingpowermodellingDVFScarbon-awarecomputingclusterresourcemanagementworkflowenergyefficiency
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper, a survey chapter rather than a new experiment, sets out to show that the carbon footprint of scientific workflows is real, quantifiable, and reducible. The authors estimate the operational emissions of three real workflows using a simple accounting formula: a geoscience run on a commodity cluster comes to about 6.95 kgCO2e per execution, an astronomy workflow on a public cloud to about 350 kgCO2e in total, and a single bioinformatics task to between 0.16 and 0.69 gCO2e, with the cloud execution about 4.5 times more energy-efficient than a decade-old cluster node. They then survey the levers that shrink these numbers, at the level of individual tasks (heterogeneous hardware, energy-aware compilers, dynamic voltage and frequency scaling, DVFS) and of whole workflows (energy-aware scheduling, consolidation, carbon-aware temporal shifting). Their concluding claim is that the need to reduce workflow emissions is clear, the techniques are numerous, yet few are widely used in practice. The upshot for a working scientist is a concrete way to account for their own workflows' emissions and a menu of practical reduction options.

What carries the argument

The mechanism that carries the paper's quantitative argument is the linear dynamic power model: total power is the sum of a static idle component and a dynamic component that rises linearly with utilisation, $P_{\mathrm{dynamic}} = (P_{\mathrm{peak}} - P_{\mathrm{static}}) \times \mathrm{utilisation}$, and emissions follow by multiplying total power by the facility's power usage effectiveness (PUE) and the grid's carbon intensity (CI). This identity turns the three worked examples into comparable numbers, and it is the same model that underlies the energy-aware scheduling literature the paper surveys. On the scheduling side, the load-bearing object is the HEFT-style two-phase heuristic — rank the workflow's tasks by their distance from the end of the task graph, then map each task to the node that best meets the objective — which GreenHEFT and MOHEFT adapt to minimise energy instead of runtime, or to trace the Pareto frontier of runtime versus energy.

What would settle it

Mount the three example workloads on instrumented hardware (Intel RAPL counters or a wall power meter at the node), record power across the full utilisation range, and compare measured energy with the model's prediction: if the error consistently exceeds the cited 14.4–47.9% band, or if direct measurement shows the cloud VM is not more efficient than the cluster node for the FastQC task, the comparative estimates would not survive. A simpler check is to re-run the FastQC comparison with the two facilities' actually measured PUE values instead of the assumed 1.6 and 1.1, and see whether the 4.5x factor holds.

Watch

Extended reading notes

Core claim

The paper's central claim is that executing scientific workflows on compute clusters causes substantial, avoidable greenhouse gas emissions, and that the means to reduce them exist at every level of the computing stack. To make the claim concrete, the authors compute ballpark footprints with the standard accounting formula $E = P_{\mathrm{total}} \times \mathrm{PUE} \times \mathrm{CI}$ — total power multiplied by the facility's power usage effectiveness and the grid's carbon intensity — where power is modelled as a static idle component plus a dynamic component that scales linearly with utilisation. Their worked examples come out at 6.95 kgCO2e for one run of a geoscience Earth-observation workflow (FORCE) on 21 nodes in Germany, around 350 kgCO2e for a 318,000-core-hour astronomy workflow on a major public cloud, and 0.16–0.69 gCO2e for a FastQC bioinformatics quality-control task, where the cloud execution is almost 4.5 times more energy-efficient than an older commodity cluster node. On the reduction side, the paper argues that task-level techniques — choosing the most energy-efficient device, compiler transformations guided by cost-performance models, dynamic voltage and frequency scaling — and workflow-level techniques — HEFT-derived energy-aware schedulers such as GreenHEFT and MOHEFT, workload consolidation, virtual-machine right-sizing, and carbon-aware shifting of delay-tolerant workflows — each meaningfully cut energy and emissions. The concluding assertion is a call to action: the need to reduce the carbon footprint of scientific workflows is clear, numerous techniques exist, and yet few are widely deployed.

Load-bearing premise

Every quantitative comparison in the paper rests on assuming that a node's power draw rises linearly with utilisation between idle and peak — the authors' own validation on a real workflow puts that model's error at 14.4% to 47.9% — so if real hardware behaves strongly non-linearly, the specific numbers, including the 4.5x ratio, could be materially wrong.

Editorial extensions

If this is right

  • A research group can estimate its own workflows' operational carbon with the paper's three-line formula, using PUE and national average carbon intensity, without needing specialised meters.
  • Delay-tolerant scientific workflows can shift execution to hours when grid carbon intensity is low, cutting emissions without changing the scientific results.
  • Choosing compute infrastructure is a real lever: the paper's FastQC comparison suggests a modern cloud VM can be roughly 4.5 times more energy-efficient than a decade-old cluster node for the same task, even after accounting for data-centre overheads.
  • Energy-aware scheduling built on the HEFT ranking-and-mapping structure, as in GreenHEFT and MOHEFT, can lower a workflow's energy consumption while respecting deadlines, provided per-task energy estimates from prior runs exist.
  • As grids decarbonise, embodied carbon — dominated by disk capacity and SSDs — overtakes operational emissions, so wasting disk space causes emissions even when the energy itself is clean.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • My inference: because the paper reports that its own linear power model deviates by 14.4–47.9% from a measured workflow, the specific numbers (6.95 kg, 350 kg, the 4.5x ratio) are directional ballparks rather than measurements; instrumenting the same three workloads with RAPL-style counters or wall meters would show how much the ratios and rankings move.
  • My inference: the formula $E = P_{\mathrm{total}} \times \mathrm{PUE} \times \mathrm{CI}$ is a general template — any workflow system that records core-hours, memory, and runtime can feed it, so the paper effectively hands every research group a carbon calculator it can run on its own traces.
  • My inference: combining carbon-aware temporal shifting with the cloud spot-market surplus the paper describes could make marginal carbon intensity a sharper scheduling signal than annual national averages, since idle but powered capacity already exists at low-carbon moments.
  • My inference: a testable extension suggested by the paper's embodied-carbon analysis is whether promptly deleting hot data and moving cold data to tape reduces a workflow's lifecycle footprint more than any code optimisation once grids are largely renewable.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

1 major / 7 minor

Summary. This chapter-style survey argues that scientific workflow execution on compute clusters has a substantial operational carbon footprint and that a variety of techniques exist to reduce it. It provides a working model of workflows and sustainable-computing concepts (Section 2), presents three worked carbon-footprint estimates for real workflows (Section 3), surveys techniques at the task level (Section 4) and workflow level (Section 5), and discusses infrastructure, energy-source, and economic considerations (Section 6). The headline qualitative conclusion is that there is a clear need and ample technical opportunity to reduce workflow emissions; the most salient quantitative sub-claims are the three estimates in Section 3, including the Section 3.4 comparison that running a FastQC task on a GCP n2 VM is almost 4.5 times more energy-efficient than on a ten-year-old TU Berlin cluster node.

Significance. If the qualitative claim holds, the paper is a useful, practitioner-oriented overview that ties together disparate literatures (compilers, scheduling, infrastructure management, carbon-aware computing) and frames them for a scientific-workflow audience. The paper makes its quantitative methodology unusually transparent: it states the linear power model, names the assumptions (50% utilisation, PUE 1.6 vs 1.1, national annual average carbon intensities), cites the sources, and even reports the 14.4-47.9% error range of its own prior validation study [WRET25]. That transparency is a genuine strength, as is the explicit call to action and the acknowledgment of rebound effects. The three worked examples are pedagogically valuable as templates for back-of-the-envelope footprint estimates, and the distinction between operational and embodied emissions is handled carefully. The survey does not claim to be exhaustive, but its coverage of task-level and workflow-level techniques is reasonable for its scope.

major comments (1)
  1. [Section 3.1] The methodology section presents Eq. (2) as a linear interpolation between static and peak power, but the paper's own background section (Section 2.2) notes that real components have non-linear power-utilisation relationships, and [WRET25] is cited as showing 14.4%-47.9% error. The paper should state more explicitly that all Section 3 estimates inherit this error and that the estimates are intentionally ballpark. Currently the error statement appears in a single sentence in Section 3.1 and is not carried through to the interpretation of the three estimates, which is inconsistent with the paper's otherwise transparent presentation.
minor comments (7)
  1. [Section 3.4] The sentence "This is further confirmed by the similar runtime..." is slightly overstrong: a similar runtime confirms similar CPU performance, not the power estimate. Suggest rephrasing to "This is consistent with the similar CPU performance...".
  2. [Section 4.4] The references [VS23a] and [VS23b] are technical reports from a scoping project, not peer-reviewed publications; the paper should state this clearly (it does in the reference list) and, ideally, summarize the key findings more substantively in the text, as the current text relies heavily on citing these reports.
  3. [Section 2.2] The definition of energy-proportional computing as "energy consumption per operation independent of utilisation" is a useful idealization, but the relation to Eq. (2) would be clearer if the paper explicitly noted that linear models imply energy proportionality only when the static power is zero; currently the background and methodology sections are not explicitly connected on this point.
  4. [Section 3.2] In the embodied emissions estimate, the lifetime share is computed as 315 minutes of a 10-year lifetime, which gives 0.00599% of lifetime usage per node; the sentence "each node was used 0.00599% of its lifetime" is correct but slightly confusing because the paper then multiplies by 21 nodes. Consider adding a phrase such as "per node" for clarity.
  5. [Section 6.3] The discussion of Jevons' paradox is brief but appropriate; the final take-away sentence ("so any savings achieved cannot mean we can run more workflows") is a normative call to action, which is fine for a chapter but should be flagged as an opinion rather than a technical conclusion.
  6. [General] The paper uses "energy consumption" and "emissions" somewhat interchangeably in the conclusion; for example, "a bioinformatics task instance can be associated with almost 4.5 times less energy consumption and operational emissions" conflates the ratio of energy (which is what was computed) with the ratio of emissions (which would be the same only when the carbon intensity is equal, which it is here, but the wording should be precise).
  7. [Figure 2] Figure 2's caption and the text in Section 4 refer to "Section 1.4.1", "Section 1.4.2", etc., which appear to be internal cross-references from a different chapter version; these should be updated to the current section numbers (4.1-4.4).

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: worked estimates are transparent arithmetic from stated power models and external coefficients; self-citations supply empirical inputs and a disclosed error caveat, not forced conclusions.

full rationale

The paper's central claims are qualitative (survey of techniques), and the quantitative estimates in Section 3 are explicitly illustrative calculations from Equations 1-3. The power figures (e.g., 34 W static / 94 W peak for the Xeon E3-1230 V2, CCF per-vCPU coefficients, PUE 1.6/1.1, national carbon intensities) come from external or disclosed sources, and the 4.5x cloud-vs-cluster ratio in Section 3.4 is just (0.91 Wh * 1.6) / (0.30 Wh * 1.1), with both energies computed by transparent linear interpolation. No parameter is fitted to the target quantity and then renamed a prediction. The self-citations are empirical: [BLT+22] provides a measured trace of runtimes and utilizations, and [WRET25] is invoked only to state a disclosed 14.4%-47.9% error caveat for this class of estimate; neither forces the conclusion. The acknowledged weakness of Equation 2's linear dynamic-power model is an accuracy/uncertainty concern, not circularity. Thus no load-bearing step reduces to its own input.

Assumptions & free parameters 4 free parameters · 4 assumptions · 0 invented entities

This is a review; it introduces no new entities. Its footprint estimates depend on several hand-chosen coefficients and domain assumptions inherited from established carbon accounting methodologies, which are the main source of uncertainty.

free parameters (4)
  • PUE for university data centre = 1.6
    Assumed from the 2021 industry average survey (Section 3.2) because the TU Berlin data centre's actual PUE is unknown. A different PUE directly scales the 6.95 kgCO2e estimate.
  • Average CPU utilisation for FORCE workflow = 50%
    Taken from the FORCE scale-out paper (Section 3.2) and used in Equation 2 to compute dynamic power. If actual utilisation differed, the energy estimate would change linearly.
  • CCF fallback power per vCPU core hour = 2.12 W
    Midpoint of the 0.74 to 3.5 W range suggested by Cloud Carbon Footprint when utilisation is unknown; used for the Montage estimate in Section 3.3.
  • National annual average carbon intensities = 439 (Germany 2021), 433 (US 2019), 473 (Germany 2022) gCO2e/kWh
    Used from ElectricityMaps in Section 3; paper acknowledges finer-grained intensities fluctuate substantially.
assumptions (4)
  • domain assumption Total node power equals static plus dynamic power, with dynamic power linear in utilisation (Equations 1-2).
    Adopted from SCI/CCF/GA methodologies in Section 3.1; underlies all estimates and much of the scheduling literature reviewed.
  • domain assumption Operational energy consumption dominates lifecycle carbon for workflow tasks under typical use.
    Stated in Section 4 to justify focusing on runtime energy; the paper itself notes embodied carbon can dominate with renewable-heavy electricity.
  • domain assumption National annual average carbon intensity is an adequate proxy for the emissions of a workflow run.
    Used throughout Section 3; the paper acknowledges that sub-national and sub-annual variation is large.
  • domain assumption Workflow tasks are black boxes connected in a directed acyclic graph.
    Standard working model of scientific workflow management systems presented in Section 2.1 and used for scheduling discussion.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Energy-Aware Workflow Execution: An Overview of Techniques for Saving Energy and Emissions in Scientific Compute Clusters." pith.science (2026). https://pith.science/paper/DQA6V2GE

@misc{pith2026250604062,
  author       = {Pith},
  title        = {Pith review of: Energy-Aware Workflow Execution: An Overview of Techniques for Saving Energy and Emissions in Scientific Compute Clusters},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DQA6V2GE}},
  note         = {Machine review of arXiv:2506.04062}
}
read the original abstract

Scientific research in many fields routinely requires the analysis of large datasets, and scientists often employ workflow systems to leverage clusters of computers for their data analysis. However, due to their size and scale, these workflow applications can have a considerable environmental footprint in terms of compute resource use, energy consumption, and carbon emissions. Mitigating this is critical in light of climate change and the urgent need to reduce carbon emissions. In this chapter, we exemplify the problem by estimating the carbon footprint of three real-world scientific workflows from different scientific domains. We then describe techniques for reducing the energy consumption and, thereby, carbon footprint of individual workflow tasks and entire workflow applications, such as using energy-efficient heterogeneous architectures, generating optimised code, scaling processor voltages and frequencies, consolidating workloads on shared cluster nodes, and scheduling workloads for optimised energy efficiency.

Figures

Figures reproduced from arXiv: 2506.04062 by the authors.

Figure 1
Figure 1. A workflow application that connects seven data processing tasks [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. An individual workflow task, divided into blocks of code, is partitioned [PITH_FULL_IMAGE:figures/full_fig_p014_2.png] view at source ↗
Figure 3
Figure 3. Task power modelling and estimation, task scheduling and placement, [PITH_FULL_IMAGE:figures/full_fig_p021_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

93 extracted references · 80 canonical work pages

  1. [1]

    Carbon explorer: A holistic framework for designing carbon aware datacenters

    Bilge Acun, Benjamin Lee, Fiodar Kazhamiaka, Kiwan Maeng, Udit Gupta, Manoj Chakkaravarthy, David Brooks, and Carole-Jean Wu. Carbon explorer: A holistic framework for designing carbon aware datacenters. In ASPLOS'23 . ACM, 2023

  2. [2]

    Energy efficient partitioning and scheduling approach for scientific workflows in the cloud

    Khadija Bousselmi, Zaki Brahmi, and Mohamed Mohsen Gammoudi. Energy efficient partitioning and scheduling approach for scientific workflows in the cloud. In SCC'16 . IEEE, 2016

  3. [3]

    Assessing ICT global emissions footprint: Trends to 2040 & recommendations

    Lotfi Belkhir and Ahmed Elmeligi. Assessing ICT global emissions footprint: Trends to 2040 & recommendations. Journal of Cleaner Production , 177, 2018

  4. [4]

    The Datacenter as a Computer: Designing Warehouse-Scale Machines

    Luiz Andr \'e Barroso, Urs H \"o lzle, and Parthasarathy Ranganathan. The Datacenter as a Computer: Designing Warehouse-Scale Machines . Springer, 2019

  5. [5]

    Using compiler techniques to improve automatic performance modeling

    Arnamoy Bhattacharyya, Grzegorz Kwasniewski, and Torsten Hoefler. Using compiler techniques to improve automatic performance modeling. In PACT'15 . IEEE, 2015

  6. [6]

    Lotaru: Locally estimating runtimes of scientific workflow tasks in heterogeneous clusters

    Jonathan Bader, Fabian Lehmann, Lauritz Thamsen, Jonathan Will, Ulf Leser, and Odej Kao. Lotaru: Locally estimating runtimes of scientific workflow tasks in heterogeneous clusters. In SSDBM'22 . ACM, 2022

  7. [7]

    Lotaru: Locally predicting workflow task runtimes for resource management on heterogeneous infrastructures

    Jonathan Bader, Fabian Lehmann, Lauritz Thamsen, Ulf Leser, and Odej Kao. Lotaru: Locally predicting workflow task runtimes for resource management on heterogeneous infrastructures. FGCS , 150, 2024

  8. [8]

    The future of FPGA acceleration in datacenters and the cloud

    Christophe Bobda, Joel Mandebi Mbongue, Paul Chow, Mohammad Ewais, Naif Tarafdar, Juan Camilo Vega, Ken Eguro, Dirk Koch, Suranga Handagala, Miriam Leeser, et al. The future of FPGA acceleration in datacenters and the cloud. ACM TRETS , 15(3), 2022

Show all 93 references
  1. [9]

    Evaluation of HPC workloads running on open-source RISC-V hardware

    Luc Berger-Vergiat, Suma G Cardwell, Ben Feinberg, Simon D Hammond, Clayton Hughes, Michael Levenhagen, and Kevin Pedretti. Evaluation of HPC workloads running on open-source RISC-V hardware. In ISC'23 . Springer, 2023

  2. [10]

    Chisel: constructing hardware in a scala embedded language

    Jonathan Bachrach, Huy Vo, Brian Richards, Yunsup Lee, Andrew Waterman, Rimas Avi z ienis, John Wawrzynek, and Krste Asanovi \'c . Chisel: constructing hardware in a scala embedded language. In DAC'12 . ACM, 2012

  3. [11]

    Evaluating energy-aware scheduling algorithms for I/O -intensive scientific workflows

    Tain \ a Coleman, Henri Casanova, Ty Gwartney, and Rafael Ferreira da Silva. Evaluating energy-aware scheduling algorithms for I/O -intensive scientific workflows. In ICCS'21 . Springer, 2021

  4. [12]

    Yang, and Peizhong Lu

    Huangke Chen, Xiaomin Zhu, Dishan Qiu, Hui Guo, Laurence T. Yang, and Peizhong Lu. EONS : Minimizing energy consumption for executing real-time workflows in virtualized cloud data centers. In ICPP-W'16 , 2016

  5. [13]

    Durillo, Vlad Nae, and Radu Prodan

    Juan J. Durillo, Vlad Nae, and Radu Prodan. Multi-objective energy-efficient workflow scheduling using list-based heuristics. FGCS , 36, 2014

  6. [14]

    Durillo, Radu Prodan, and Jorge G

    Juan J. Durillo, Radu Prodan, and Jorge G. Barbosa. Pareto tradeoff scheduling of workflows on federated commercial clouds. Simulation Modelling Practice and Theory , 58, 2015

  7. [15]

    Pegasus in the cloud: Science automation through workflow technologies

    Ewa Deelman, Karan Vahi, Mats Rynge, Gideon Juve, Rajiv Mayani, and Rafael Ferreira da Silva. Pegasus in the cloud: Science automation through workflow technologies. IEEE Internet Computing , 20(1), 2016

  8. [16]

    Blair, and Adrian Friday

    Charlotte Freitag, Mike Berners-Lee, Kelly Widdicks, Bran Knowles, Gordon S. Blair, and Adrian Friday. The real climate and transformative impact of ICT : A critique of estimates, trends, and regulations. Patterns , 2(9), 2021

  9. [17]

    Characterizing, modeling, and accurately simulating power and energy consumption of I/O -intensive scientific workflows

    Rafael Ferreira da Silva , Henri Casanova, Anne-Cécile Orgerie, Ryan Tanaka, Ewa Deelman, and Frédéric Suter. Characterizing, modeling, and accurately simulating power and energy consumption of I/O -intensive scientific workflows. Journal of Computational Science , 44, 2020

  10. [18]

    Accurately simulating energy consumption of I/O -intensive scientific workflows

    Rafael Ferreira da Silva, Anne-C \'e cile Orgerie, Henri Casanova, Ryan Tanaka, Ewa Deelman, and Fr \'e d \'e ric Suter. Accurately simulating energy consumption of I/O -intensive scientific workflows. In ICCS'19 . Springer, 2019

  11. [19]

    Force: Landsat + sentinel-2 analysis ready data and beyond

    David Frantz. Force: Landsat + sentinel-2 analysis ready data and beyond. Remote Sensing , 11(9), 2019

  12. [20]

    Lee, Gu-Yeon Wei, David Brooks, and Carole-Jean Wu

    Udit Gupta, Young Geun Kim, Sylvia Lee, Jordan Tse, Hsien-Hsin S. Lee, Gu-Yeon Wei, David Brooks, and Carole-Jean Wu. Chasing carbon: The elusive environmental footprint of computing. In HPCA'21 . IEEE, 2021

  13. [21]

    Energy efficient scheduling of scientific workflows in cloud environment

    Manojit Ghose, Pratyush Verma, Sushanta Karmakar, and Aryabartta Sahu. Energy efficient scheduling of scientific workflows in cloud environment. In HPCC/SmartCity/DSS'17 . IEEE, 2017

  14. [22]

    Measurement and characterization of haswell power and energy consumption

    Song Huang, Michael Lang, Scott Pakin, and Song Fu. Measurement and characterization of haswell power and energy consumption. In E2SC '15 , E2SC '15. ACM, 2015

  15. [23]

    Reducing cloud expenditures and carbon emissions via virtual machine migration and downsizing

    Nathan Huang, Anthony Li, Sophia Zhang, and Ziliang Zong. Reducing cloud expenditures and carbon emissions via virtual machine migration and downsizing. In IPCCC'23 . IEEE, 2023

  16. [24]

    A review of power consumption models of servers in data centers

    Chaoqiang Jin, Xuelian Bai, Chao Yang, Wangxin Mao, and Xin Xu. A review of power consumption models of servers in data centers. Applied Energy , 265, 2020

  17. [25]

    Performance comparision of TPU , GPU , CPU on google colaboratory over distributed deep learning

    Haklin Kimm, Incheon Paik, and Hanke Kimm. Performance comparision of TPU , GPU , CPU on google colaboratory over distributed deep learning. In MCSoC'21 . IEEE, 2021

  18. [26]

    Energy-efficient task scheduling and consolidation algorithm for workflow jobs in cloud

    Mustafa Khaleel and Michelle M Zhu. Energy-efficient task scheduling and consolidation algorithm for workflow jobs in cloud. International Journal of Computational Science and Engineering , 13(3), 2016

  19. [27]

    Mlir: Scaling compiler infrastructure for domain specific computation

    Chris Lattner, Mehdi Amini, Uday Bondhugula, Albert Cohen, Andy Davis, Jacques Pienaar, River Riddle, Tatiana Shpeisman, Nicolas Vasilache, and Oleksandr Zinenko. Mlir: Scaling compiler infrastructure for domain specific computation. In CGO'21 . IEEE, 2021

  20. [28]

    Survey of energy-efficient and power-proportional storage systems

    Pablo Llopis, Javier Garcia Blas, Florin Isaila, and Jesus Carretero. Survey of energy-efficient and power-proportional storage systems. The Computer J. , 57(7), 2013

  21. [29]

    How Workflow Engines Should Talk to Resource Managers: A Proposal for a Common Workflow Scheduling Interface

    Fabian Lehmann, Jonathan Bader, Friedrich Tschirpke, Lauritz Thamsen, and Ulf Leser. How Workflow Engines Should Talk to Resource Managers: A Proposal for a Common Workflow Scheduling Interface . In CCGrid'23 . IEEE, 2023

  22. [30]

    Invited paper: A compile-time cost model for OpenMP

    Chunhua Liao and Barbara Chapman. Invited paper: A compile-time cost model for OpenMP . In IPDPS'07 . IEEE, 2007

  23. [31]

    Adapting datacenter capacity for greener datacenters and grid

    Liuzixuan Lin and Andrew A Chien. Adapting datacenter capacity for greener datacenters and grid. In e-Energy'23 . ACM, 2023

  24. [32]

    Heterohalide: From image processing DSL to efficient FPGA acceleration

    Jiajie Li, Yuze Chi, and Jason Cong. Heterohalide: From image processing DSL to efficient FPGA acceleration. In FPGA'20 . ACM, 2020

  25. [33]

    Force on nextflow: Scalable analysis of earth observation data on commodity clusters

    Fabian Lehmann, David Frantz, Sören Becker, Ulf Leser, and Patrick Hostert. Force on nextflow: Scalable analysis of earth observation data on commodity clusters. In CIKM-W'21 . CEUR-WS.org , 2021

  26. [34]

    Cost and energy aware scheduling algorithm for scientific workflows with deadline constraint in clouds

    Zhongjin Li, Jidong Ge, Haiyang Hu, Wei Song, Hao Hu, and Bin Luo. Cost and energy aware scheduling algorithm for scientific workflows with deadline constraint in clouds. IEEE Transactions on Services Computing , 11(4), 2018

  27. [35]

    Green algorithms: Quantifying the carbon footprint of computation

    Loïc Lannelongue, Jason Grealey, and Michael Inouye. Green algorithms: Quantifying the carbon footprint of computation. Adv. Science , 8(12), 2021

  28. [36]

    Spotlake: Diverse spot instance dataset archive service

    Sungjae Lee, Jaeil Hwang, and Kyungyong Lee. Spotlake: Diverse spot instance dataset archive service. In IISWC'22 . IEEE, 2022

  29. [37]

    An experimental study of data retention behavior in modern dram devices: Implications for retention time profiling mechanisms

    Jamie Liu, Ben Jaiyen, Yoongu Kim, Chris Wilkerson, and Onur Mutlu. An experimental study of data retention behavior in modern dram devices: Implications for retention time profiling mechanisms. In ISCA'13 . ACM, 2013

  30. [38]

    Heterogeneous systems for energy efficient scientific computing

    Qiang Liu and Wayne Luk. Heterogeneous systems for energy efficient scientific computing. In Oliver C. S. Choy, Ray C. C. Cheung, Peter Athanas, and Kentaro Sano, editors, Reconfigurable Computing: Architectures, Tools and Applications . Springer, 2012

  31. [39]

    Dynamic voltage and frequency scaling: The laws of diminishing returns

    Etienne Le Sueur and Gernot Heiser. Dynamic voltage and frequency scaling: The laws of diminishing returns. In HotPower'10 . USENIX, 2010

  32. [40]

    u rthinger, and Hanspeter M\

    David Leopoldseder, Lukas Stadler, Manuel Rigger, Thomas W\" u rthinger, and Hanspeter M\" o ssenb\" o ck. A cost model for a graph-based intermediate-representation in a dynamic compiler. In MIL'18 . ACM, 2018

  33. [41]

    Compoff: A compiler cost model using machine learning to predict the cost of OpenMP offloading

    Alok Mishra, Smeet Chheda, Carlos Soto, Abid M Malik, Meifeng Lin, and Barbara Chapman. Compoff: A compiler cost model using machine learning to predict the cost of OpenMP offloading. In IPDPS-W'22 . IEEE, 2022

  34. [42]

    WattsApp : Power-aware container scheduling

    Hemant Kumar Mehta, Paul Harvey, Omer Rana, Rajkumar Buyya, and Blesson Varghese. WattsApp : Power-aware container scheduling. In UCC'20 . IEEE, 2020

  35. [43]

    Energy efficient workflow scheduling with virtual machine consolidation for green cloud computing

    N Mohanapriya, G Kousalya, P Balakrishnan, and C Pethuru Raj. Energy efficient workflow scheduling with virtual machine consolidation for green cloud computing. Journal of Intelligent & Fuzzy Systems , 34(3), 2018

  36. [44]

    Bringing carbon awareness to multi-cloud application delivery

    Diptyaroop Maji, Ben Pfaff, Vipin P R, Rajagopal Sreenivasan, Victor Firoiu, Sreeram Iyer, Colleen Josephson, Zhelong Pan, and Ramesh K Sitaraman. Bringing carbon awareness to multi-cloud application delivery. In HotCarbon'23 . ACM, 2023

  37. [45]

    Energy-aware workflow task scheduling in clouds with virtual machine consolidation using discrete water wave optimization

    Rambabu Medara, Ravi Shankar Singh, and Amit. Energy-aware workflow task scheduling in clouds with virtual machine consolidation using discrete water wave optimization. Simulation Modelling Practice and Theory , 110, 2021

  38. [46]

    Recalibrating global data center energy-use estimates

    Eric Masanet, Arman Shehabi, Nuoa Lei, Sarah Smith, and Jonathan Koomey. Recalibrating global data center energy-use estimates. Science , 367(6481), 2020

  39. [47]

    A performance analysis of the first generation of HPC -optimized A rm processors

    Simon McIntosh-Smith, James Price, Tom Deakin, and Andrei Poenaru. A performance analysis of the first generation of HPC -optimized A rm processors. Concurrency and Computation: Practice and Experience , 31(16), 2019

  40. [48]

    Sitaraman

    Diptyaroop Maji, Prashant Shenoy, and Ramesh K. Sitaraman. Carboncast: Multi-day forecasting of grid carbon intensity. In BuildSys'22 . ACM, 2022

  41. [49]

    Pablo Antonio Mart\' nez, Jackson Woodruff, Jordi Armengol-Estap\' e , Gregorio Bernab\' e , Jos\' e Manuel Garc\' a, and Michael F. P. O'Boyle. Matching linear algebra and tensor code to specialized hardware accelerators. In CC'23 . ACM, 2023

  42. [50]

    Mp-stream: A memory performance benchmark for design space exploration on heterogeneous HPC devices

    Syed Waqar Nabi and Wim Vanderbauwhede. Mp-stream: A memory performance benchmark for design space exploration on heterogeneous HPC devices. In IPDPS-W'18 . IEEE, 2018

  43. [51]

    FPGA design space exploration for scientific HPC applications using a fast and accurate cost model based on roofline analysis

    Syed Waqar Nabi and Wim Vanderbauwhede. FPGA design space exploration for scientific HPC applications using a fast and accurate cost model based on roofline analysis. Journal of Parallel and Distributed Computing , 133, 2019

  44. [52]

    Energy-constrained provisioning for scientific workflow ensembles

    Ilia Pietri, Maciej Malawski, Gideon Juve, Ewa Deelman, Jarek Nabrzyski, and Rizos Sakellariou. Energy-constrained provisioning for scientific workflow ensembles. In CGC'13 , 2013

  45. [53]

    Deep learning with tensorflow: A review

    Bo Pang, Erik Nijkamp, and Ying Nian Wu. Deep learning with tensorflow: A review. Journal of Educational and Behavioral Statistics , 45(2), 2020

  46. [54]

    Producing an infrared multiwavelength galactic plane atlas using montage, pegasus and amazon web services

    Mats Rynge, Gideon Juve, Jamie Kinney, John Good, Bruce Berriman, Ann Merrihew, and Ewa Deelman. Producing an infrared multiwavelength galactic plane atlas using montage, pegasus and amazon web services. In ADASS'13 . The Astronomical Society of the Pacific, 2013

  47. [55]

    Carbon-aware computing for datacenters

    Ana Radovanović, Ross Koningstein, Ian Schneider, Bokan Chen, Alexandre Duarte, Binz Roy, Diyue Xiao, Maya Haridasan, Patrick Hung, Nick Care, Saurav Talukdar, Eric Mullen, Kendal Smith, MariEllen Cottman, and Walfredo Cirne. Carbon-aware computing for datacenters. IEEE Transa...

  48. [56]

    A multi-objective based scheduling framework for effective resource utilization in cloud computing

    Pillareddy Vamsheedhar Reddy and Karri Ganesh Reddy. A multi-objective based scheduling framework for effective resource utilization in cloud computing. IEEE Access , 11, 2023

  49. [57]

    The Mont-Blanc prototype: an alternative approach for HPC systems

    Nikola Rajovic, Alejandro Rico, Filippo Mantovani, Daniel Ruiz, Josep Oriol Vilarrubi, Constantino Gomez, Luna Backes, Diego Nieto, Harald Servat, Xavier Martorell, et al. The Mont-Blanc prototype: an alternative approach for HPC systems. In SC'16 . ACM, 2016

  50. [58]

    Energy aware scheduling algorithms for cloud environments -- a survey

    Sumit Kumar Saurav and Shajulin Benedict. Energy aware scheduling algorithms for cloud environments -- a survey. In ACCESS'21 . IEEE, 2021

  51. [59]

    Ecovisor: A virtual energy system for carbon-efficient applications

    Abel Souza, Noman Bashir, Jorge Murillo, Walid Hanafy, Qianlin Liang, David Irwin, and Prashant Shenoy. Ecovisor: A virtual energy system for carbon-efficient applications. In ASPLOS'23 . ACM, 2023

  52. [60]

    Vishakha Singh, Indrajeet Gupta, and Prasanta K. Jana. An energy efficient algorithm for workflow scheduling in IaaS cloud. J. Grid Comput. , 18(3), 2020

  53. [61]

    Memory-aware functional ir for higher-level synthesis of accelerators

    Christof Schlaak, Tzung-Han Juang, and Christophe Dubach. Memory-aware functional ir for higher-level synthesis of accelerators. ACM TACO , 19(2), 2022

  54. [62]

    Stavrinides and Helen D

    Georgios L. Stavrinides and Helen D. Karatza. An energy-efficient, qos-aware and cost-effective scheduling approach for real-time workflow applications in cloud computing systems utilizing dvfs and approximate computations. FGCS , 96, 2019

  55. [63]

    Reducing FPGA memory footprint of stencil codes through automatic extraction of memory patterns

    Robert Szafarczyk, Syed Waqar Nabi, and Wim Vanderbauwhede. Reducing FPGA memory footprint of stencil codes through automatic extraction of memory patterns. In FPL'22 . IEEE, 2022

  56. [64]

    Lift: a functional data-parallel ir for high-performance GPU code generation

    Michel Steuwer, Toomas Remmelg, and Christophe Dubach. Lift: a functional data-parallel ir for high-performance GPU code generation. In CGO'17 . IEEE, 2017

  57. [65]

    Obsidian: A domain specific embedded language for parallel programming of graphics processors

    Joel Svensson, Mary Sheeran, and Koen Claessen. Obsidian: A domain specific embedded language for parallel programming of graphics processors. In IFL'08 . Springer, 2008

  58. [66]

    PASTA : a power-aware solution to scheduling of precedence-constrained tasks on heterogeneous computing resources

    Mohsen Sharifi, Saeed Shahrivari, and Hadi Salimi. PASTA : a power-aware solution to scheduling of precedence-constrained tasks on heterogeneous computing resources. Computing , 95(1), 2013

  59. [67]

    A design space exploration of compiler optimizations guided by hot functions

    Marcos Yukio Siraichi, Caio Tonetti, and Anderson Faustino da Silva. A design space exploration of compiler optimizations guided by hot functions. In SCCC'16 . IEEE, 2016

  60. [68]

    Static analysis: An introduction: The fundamental challenge of software engineering is one of complexity

    Patrick Thomson. Static analysis: An introduction: The fundamental challenge of software engineering is one of complexity. Queue , 19(4), 2021

  61. [69]

    Performance-effective and low-complexity task scheduling for heterogeneous computing

    Haluk Topcuoglu, Salim Hariri, and Min-You Wu. Performance-effective and low-complexity task scheduling for heterogeneous computing. IEEE TPDS , 13(3), 2002

  62. [70]

    The dirty secret of ssds: Embodied carbon

    Swamit Tannu and Prashant J Nair. The dirty secret of ssds: Embodied carbon. ACM SIGENERGY Energy Informatics Review , 3(3), 2023

  63. [71]

    Polyhedral-model guided loop-nest auto-vectorization

    Konrad Trifunovic, Dorit Nuzman, Albert Cohen, Ayal Zaks, and Ira Rosen. Polyhedral-model guided loop-nest auto-vectorization. In PACT'09 . IEEE, 2009

  64. [72]

    Efficient FPGA cost-performance space exploration using type-driven program transformations

    Cristian Urlea, Wim Vanderbauwhede, and Syed Waqar Nabi. Efficient FPGA cost-performance space exploration using type-driven program transformations. In ReConFig'19 . IEEE, 2019

  65. [73]

    Vanderbauwhede, L

    W. Vanderbauwhede, L. Azzopardi, and M. Moadeli. FPGA -accelerated information retrieval: High-efficiency document filtering. In FPL'09 . IEEE, 2009

  66. [74]

    Meta-programming design-flow patterns for automating reusable optimisations

    Jessica Vandebon, Jose Coutinho, and Wayne Luk. Meta-programming design-flow patterns for automating reusable optimisations. In HEART'22 . ACM, 2022

  67. [75]

    Domain-specific acceleration and auto-parallelization of legacy scientific code in fortran 77 using source-to-source compilation

    Wim Vanderbauwhede and Gavin Davidson. Domain-specific acceleration and auto-parallelization of legacy scientific code in fortran 77 using source-to-source compilation. Computers & Fluids , 173, 2018

  68. [76]

    Type-driven automated program transformations and cost modelling for optimising streaming programs on FPGA s

    Wim Vanderbauwhede, Syed Waqar Nabi, and Cristian Urlea. Type-driven automated program transformations and cost modelling for optimising streaming programs on FPGA s. International Journal of Parallel Programming , 47, 2019

  69. [77]

    Impact of the development and deployment of software on energy consumption: findings and recommendations , 2023

    Wim Vanderbauwhede and Ole Stubben. Impact of the development and deployment of software on energy consumption: findings and recommendations , 2023. Part of the UKRI Net Zero Digital Research Infrastructure Scoping Project

  70. [78]

    Literature review of approaches to improve software energy efficiency: findings and recommendations , 2023

    Wim Vanderbauwhede and Ole Stubben. Literature review of approaches to improve software energy efficiency: findings and recommendations , 2023. Part of the UKRI Net Zero Digital Research Infrastructure Scoping Project

  71. [79]

    An investigation into the feasibility and benefits of GPU /multicore acceleration of the weather research and forecasting model

    Wim Vanderbauwhede and Tetsuya Takemi. An investigation into the feasibility and benefits of GPU /multicore acceleration of the weather research and forecasting model. In HPCS'13 . IEEE, 2013

  72. [80]

    Let's wait awhile: How temporal workload shifting can reduce carbon emissions in the cloud

    Philipp Wiesner, Ilja Behnke, Dominik Scheinert, Kordian Gontarska, and Lauritz Thamsen. Let's wait awhile: How temporal workload shifting can reduce carbon emissions in the cloud. In Middleware'21 . ACM , 2021

  73. [81]

    Compiler-driven dependence profiling to guide program parallelization

    Peng Wu, Arun Kejariwal, and C a lin Ca s caval. Compiler-driven dependence profiling to guide program parallelization. In Jos \'e Nelson Amaral, editor, Languages and Compilers for Parallel Computing . Springer, 2008

  74. [82]

    Fedzero: Leveraging renewable excess energy in federated learning

    Philipp Wiesner, Ramin Khalili, Dennis Grinwald, Pratik Agrawal, Lauritz Thamsen, and Odej Kao. Fedzero: Leveraging renewable excess energy in federated learning. In e-Energy'24 . ACM, 2024

  75. [83]

    Exploring the Potential of Carbon-Aware Execution for Scientific Workflows

    Kathleen West, Fabian Lehmann, Vasilis Bountris, Ulf Leser, Yehia Elkhatib, and Lauritz Thamsen. Exploring the Potential of Carbon-Aware Execution for Scientific Workflows . In CCGrid'25 . IEEE, 2025

  76. [84]

    Adopting GPU computing to support DL -based earth science applications

    Zifu Wang, Yun Li, Kevin Wang, Jacob Cain, Mary Salami, Daniel Q Duffy, Michael M Little, and Chaowei Yang. Adopting GPU computing to support DL -based earth science applications. International Journal of Digital Earth , 16(1), 2023

  77. [85]

    Ichnos: A carbon footprint estimator for scientific workflows

    Kathleen West, Magnus Reid, Yehia Elkhatib, and Lauritz Thamsen. Ichnos: A carbon footprint estimator for scientific workflows. In LOCO'24 . arXiv proceedings, 2025

  78. [86]

    Cucumber: Renewable-aware admission control for delay-tolerant cloud and edge workloads

    Philipp Wiesner, Dominik Scheinert, Thorsten Wittkopp, Lauritz Thamsen, and Odej Kao. Cucumber: Renewable-aware admission control for delay-tolerant cloud and edge workloads. In Euro-Par'22 . Springer, 2022

  79. [87]

    Scheduling for reduced CPU energy

    Mark Weiser, Brent Welch, Alan Demers, and Scott Shenker. Scheduling for reduced CPU energy. Mobile Computing , 1996

  80. [88]

    EnReal : An energy-aware resource allocation method for scientific workflow executions in cloud environment

    Xiaolong Xu, Wanchun Dou, Xuyun Zhang, and Jinjun Chen. EnReal : An energy-aware resource allocation method for scientific workflow executions in cloud environment. IEEE Transactions on Cloud Computing , 4(2), 2015

  81. [89]

    Multi-objective approach for energy-aware workflow scheduling in cloud computing environments

    Sonia Yassa, Rachid Chelouah, Hubert Kadima, Bertrand Granado, et al. Multi-objective approach for energy-aware workflow scheduling in cloud computing environments. The Scientific World Journal , 2013, 2013

  82. [90]

    Chien, and Sangwon Suh

    Jiajia Zheng, Andrew A. Chien, and Sangwon Suh. Mitigating curtailment and carbon emissions through load migration between data centers. Joule , 4(10), 2020

  83. [91]

    Program locality analysis using reuse distance

    Yutao Zhong, Xipeng Shen, and Chen Ding. Program locality analysis using reuse distance. ACM TOPLAS , 31(6), 2009

  84. [92]

    Minimizing energy consumption scheduling algorithm of workflows with cost budget constraint on heterogeneous cloud computing systems

    Longxin Zhang, Lan Wang, Zhicheng Wen, Mansheng Xiao, and Junfeng Man. Minimizing energy consumption scheduling algorithm of workflows with cost budget constraint on heterogeneous cloud computing systems. IEEE Access , 8, 2020

  85. [93]

    Power-aware consolidation of scientific workflows in virtualized environments

    Qian Zhu, Jiedan Zhu, and Gagan Agrawal. Power-aware consolidation of scientific workflows in virtualized environments. In SC'10 . ACM, 2010

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.