pith. sign in

arxiv: 2606.25540 · v1 · pith:2GD6IESUnew · submitted 2026-06-24 · 💻 cs.NI · math.OC

Dependency-Aware Dominant Resource Fairness for Multi-Tenant Multi-Resource Systems

Braik Zeidan (Cnam , CEDRIC - ROC) , Francesca Fossati (CEDRIC) , Sahar Hoteit (L2S , IUF) , Stefano Secci (CEDRIC - ROC) This is my paper

Pith reviewed 2026-06-25 20:11 UTC · model grok-4.3

classification 💻 cs.NI math.OC
keywords dominant resource fairnessmulti-resource allocationdependency-aware schedulingPareto efficiencymulti-tenant systemsresource wastevRANJain fairness index
0
0 comments X

The pith

DDRF equalizes active dominant shares of congested resources to guarantee saturation of at least one resource and eliminate waste.

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

The paper proposes Dependency-aware Dominant Resource Fairness (DDRF) to handle multi-resource allocation when resources have fixed proportional dependencies and demand exceeds capacity. Standard DRF can leave resources allocated but unused by low-demand tenants because it ignores those dependencies. DDRF restricts equalization to the active dominant shares of only the currently congested resources. This produces allocations that the authors prove always fill at least one congested resource completely. The resulting policy keeps the fairness properties of DRF while removing the waste observed in cloud and virtualized radio access network traces.

Core claim

DDRF equalizes the active dominant shares of congested resources and we prove that this always saturates at least one congested resource, thereby guaranteeing Pareto efficiency and zero resource waste under the assumed linear dependency model.

What carries the argument

Active-dominant-share equalization, which computes tenant allocations by considering only the dominant shares among currently congested resources and their known linear dependency relations.

If this is right

  • DDRF always saturates at least one congested resource, ensuring Pareto efficiency.
  • Effective user satisfaction rises by up to 80 percent compared with dependency-agnostic baselines.
  • Resource waste falls by up to 60 percent relative to the same baselines.
  • Jain's fairness index improves by more than 15 percent over a purely utilitarian allocation.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • If dependencies can be measured or learned at runtime, the same equalization logic could be applied without assuming they are known statically.
  • The saturation guarantee might extend to systems where only approximate or partial dependency information is available.
  • In settings without a central orchestrator, a distributed version of active-dominant-share equalization could still reduce waste if local views of congestion are consistent.

Load-bearing premise

Inter-resource dependencies are known in advance, fixed, and can be expressed as linear proportions that allow central computation of active dominant shares without new overhead.

What would settle it

Apply DDRF to a system with measured linear dependencies and check whether any resulting allocation leaves every congested resource below full utilization while total demand exceeds capacity.

Figures

Figures reproduced from arXiv: 2606.25540 by Braik Zeidan (Cnam, CEDRIC - ROC), Francesca Fossati (CEDRIC), IUF), Sahar Hoteit (L2S, Stefano Secci (CEDRIC - ROC).

Figure 1
Figure 1. Figure 1: Comparison of dependency-aware and dependency-agnostic alloca [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: DRF inefficiency in demand satisfaction under congestion: allocation [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: A system model example with 4 tenants, 3 InPs, 4 resources. [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Average partitioning of total resource capacity into wasted, useful, and idle components under different dependency structures across congestion [PITH_FULL_IMAGE:figures/full_fig_p011_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: CDF of effective overall satisfaction rate across users and resources, aggregated over all congestion profiles. [PITH_FULL_IMAGE:figures/full_fig_p011_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: CDF of users’ minimum effective satisfaction rate across resources, aggregated over all congestion profiles. [PITH_FULL_IMAGE:figures/full_fig_p011_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Boxplots of Jain’s fairness index across congestion profiles for DDRF [PITH_FULL_IMAGE:figures/full_fig_p011_7.png] view at source ↗
Figure 10
Figure 10. Figure 10: Boxplots of Jain’s fairness index across congestion profiles for DDRF [PITH_FULL_IMAGE:figures/full_fig_p012_10.png] view at source ↗
Figure 9
Figure 9. Figure 9: CDF of effective (a) overall and (b) minimum satisfaction rate across [PITH_FULL_IMAGE:figures/full_fig_p012_9.png] view at source ↗
read the original abstract

Multi-resource allocation in network-congested, multi-tenant systems in which demand exceeds available capacity is challenging, as there is no straightforward way to determine how much of each resource to assign, especially when resources are interdependent. Classical approaches such as Dominant Resource Fairness (DRF), which generalizes Max-Min Fairness (MMF) to multiple resources, assume linear proportional dependencies across resources, requiring allocations to follow fixed proportions implied by tenants demands. However, this assumption may lead to inefficient allocations and resource waste, with allocated resources that go unused in practice. In this paper, we consider a multi-resource orchestrator and propose the Dependency-aware Dominant Resource Fairness (DDRF) policy, a centralized generalization of DRF that considers inter-resource dependencies: it equalizes active dominant shares of congested resources, preserving DRFs desirable properties, while avoiding its inefficiency with low-demand tenants. We prove that DDRF always saturates at least one congested resource, ensuring Pareto efficiency and eliminating resource waste. We evaluate DDRF using Amazon EC2 traces and a virtualized radio access network (vRAN) use case while considering real resource dependencies. The results show that DDRF improves effective user satisfaction by up to 80% and reduces resource waste by up to 60% compared to dependency-agnostic baselines, while improving Jain's fairness index by more than 15% compared to the utilitarian policy.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The manuscript introduces Dependency-Aware Dominant Resource Fairness (DDRF), a centralized generalization of Dominant Resource Fairness (DRF) for multi-tenant multi-resource systems with inter-resource dependencies. DDRF equalizes active dominant shares of congested resources, proves that it always saturates at least one congested resource (ensuring Pareto efficiency and eliminating waste), and reports trace-driven results on Amazon EC2 traces and a vRAN use case showing up to 80% higher effective user satisfaction, up to 60% lower resource waste, and >15% better Jain's fairness index versus dependency-agnostic baselines.

Significance. If the saturation proof holds under the paper's modeling assumptions, DDRF supplies a fairness mechanism that removes the waste induced by fixed-proportion DRF allocations while retaining its core properties. The magnitude of the reported gains on real traces and the vRAN case study indicates direct relevance to cloud and network orchestration, where dependent resources are common.

major comments (2)
  1. [Saturation proof section] Saturation proof (the section containing the claim that DDRF always saturates at least one congested resource): the argument relies on inter-resource dependencies being known exactly, fixed, and expressible as linear/proportional relations that permit exact central computation of the active-dominant-share equalization vector. No analysis is given for estimation error, runtime measurement, or time-varying dependencies; under those conditions the equalization step can produce allocations that leave all congested resources unsaturated, directly contradicting the Pareto-efficiency guarantee.
  2. [§5] Evaluation (§5, trace-driven experiments): the reported improvements (80% satisfaction, 60% waste reduction) are presented without explicit statement of data-exclusion rules, dependency-extraction method from the traces, or sensitivity to those choices; this leaves open whether post-hoc selection affects the cross-policy comparison.
minor comments (1)
  1. [Model section] The definition of 'active dominant share' and the precise mapping from dependency relations to the equalization step should be stated with an equation in the model section to make the extension from DRF fully self-contained.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments. We address each major comment below and indicate the revisions we will make.

read point-by-point responses

  1. Referee: [Saturation proof section] Saturation proof (the section containing the claim that DDRF always saturates at least one congested resource): the argument relies on inter-resource dependencies being known exactly, fixed, and expressible as linear/proportional relations that permit exact central computation of the active-dominant-share equalization vector. No analysis is given for estimation error, runtime measurement, or time-varying dependencies; under those conditions the equalization step can produce allocations that leave all congested resources unsaturated, directly contradicting the Pareto-efficiency guarantee.

    Authors: We agree that the saturation proof is derived under the modeling assumption of exact, fixed, and linear inter-resource dependencies known precisely to the central allocator. The manuscript provides no analysis of estimation error, runtime measurement, or time-varying dependencies, and such conditions could indeed invalidate the saturation guarantee. We will revise the proof section to state these assumptions explicitly and add a limitations paragraph clarifying that the Pareto-efficiency claim holds only when the assumptions are satisfied. revision: yes

  2. Referee: [§5] Evaluation (§5, trace-driven experiments): the reported improvements (80% satisfaction, 60% waste reduction) are presented without explicit statement of data-exclusion rules, dependency-extraction method from the traces, or sensitivity to those choices; this leaves open whether post-hoc selection affects the cross-policy comparison.

    Authors: We concur that the evaluation would be strengthened by greater methodological transparency. In the revised manuscript we will insert explicit descriptions of the dependency-extraction procedure applied to the Amazon EC2 traces, any data-exclusion criteria used, and sensitivity results with respect to those choices. revision: yes

Circularity Check

0 steps flagged

No circularity; DDRF saturation proof is self-contained under model assumptions

full rationale

The paper defines DDRF as a centralized generalization of DRF that equalizes active dominant shares of congested resources while incorporating known inter-resource dependencies. The claimed proof that DDRF always saturates at least one congested resource (ensuring Pareto efficiency) follows directly from this equalization rule under the stated assumptions of fixed, known linear/proportional dependencies. No equation or step reduces by construction to a fitted parameter, self-referential definition, or load-bearing self-citation. Evaluation uses external traces (Amazon EC2, vRAN) rather than internal fits. The derivation chain remains independent of its inputs; the strong modeling assumptions affect correctness but do not create circularity.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on standard fairness axioms and the assumption that dependencies are known and fixed; no free parameters or invented entities are introduced in the abstract.

axioms (1)
  • domain assumption Inter-resource dependencies can be expressed such that active dominant shares can be equalized centrally.
    Invoked to define the DDRF allocation rule and prove saturation.

pith-pipeline@v0.9.1-grok · 5813 in / 1157 out tokens · 17420 ms · 2026-06-25T20:11:39.708507+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

41 extracted references · 3 canonical work pages · 1 internal anchor

  1. [1]

    Surplus fair scheduling: A proportional-share cpu scheduling algorithm for symmetric multipro- cessors,

    A. Chandra, M. Adler, P. Goyal, and P. Shenoy, “Surplus fair scheduling: A proportional-share cpu scheduling algorithm for symmetric multipro- cessors,” in Proceedings of the 4th Symposium on Operating Systems Design and Implementation (OSDI) . USENIX Association, 2000

    2000

  2. [2]

    Apollo: Scalable and coordinated scheduling for cloud- scale computing,

    E. Boutin, J. Ekanayake, W. Lin, B. Shi, J. Zhou, Z. Qian, M. Wu, and L. Zhou, “Apollo: Scalable and coordinated scheduling for cloud- scale computing,” in Proceedings of the 11th USENIX Symposium on Operating Systems Design and Implementation (OSDI) . USENIX Association, 2014, pp. 285–300

    2014

  3. [3]

    Network slicing to enable scalability and flexibility in 5g mobile networks,

    P. Rost, C. Mannweiler, D. S. Michalopoulos, C. Sartori, V . Sciancale- pore, N. Sastry, O. Holland, S. Tayade, B. Han, D. Bega, D. Aziz, and H. Bakker, “Network slicing to enable scalability and flexibility in 5g mobile networks,” IEEE Communications Magazine , vol. 55, no. 5, pp. 72–79, 2017

    2017

  4. [4]

    Charging and rate control for elastic traffic,

    F. Kelly, “Charging and rate control for elastic traffic,” European transactions on Telecommunications, vol. 8, no. 1, pp. 33–37, 1997

    1997

  5. [5]

    Rate control for com- munication networks: shadow prices, proportional fairness and stability,

    F. P. Kelly, A. K. Maulloo, and D. K. H. Tan, “Rate control for com- munication networks: shadow prices, proportional fairness and stability,” Journal of the Operational Research society, vol. 49, no. 3, pp. 237–252, 1998

    1998

  6. [6]

    Bertsekas and R

    D. Bertsekas and R. Gallager, Data networks. Athena Scientific, 2021

    2021

  7. [7]

    Fair end-to-end window-based congestion control,

    J. Mo and J. Walrand, “Fair end-to-end window-based congestion control,” IEEE/ACM Transactions on networking, vol. 8, no. 5, pp. 556– 567, 2002

    2002

  8. [8]

    T. Lan, D. Kao, M. Chiang, and A. Sabharwal, An axiomatic theory of fairness in network resource allocation . IEEE, 2010

    2010

  9. [9]

    Dominant resource fairness: Fair allocation of multiple resource types,

    A. Ghodsi, M. Zaharia, B. Hindman, A. Konwinski, S. Shenker, and I. Stoica, “Dominant resource fairness: Fair allocation of multiple resource types,” in 8th USENIX symposium on networked systems design and implementation (NSDI 11) , 2011

    2011

  10. [10]

    Multi-resource fair allocation in heterogeneous cloud computing systems,

    W. Wang, B. Liang, and B. Li, “Multi-resource fair allocation in heterogeneous cloud computing systems,”IEEE Transactions on Parallel and Distributed Systems , vol. 26, no. 10, pp. 2822–2835, 2014

    2014

  11. [11]

    Multiresource allocation: Fairness–efficiency tradeoffs in a unifying framework,

    C. Joe-Wong, S. Sen, T. Lan, and M. Chiang, “Multiresource allocation: Fairness–efficiency tradeoffs in a unifying framework,” IEEE/ACM Transactions on Networking , vol. 21, no. 6, pp. 1785–1798, 2013

    2013

  12. [12]

    Multi-resource allocation for network slicing,

    F. Fossati, S. Moretti, P. Perny, and S. Secci, “Multi-resource allocation for network slicing,” IEEE/ACM Transactions on Networking , vol. 28, no. 3, pp. 1311–1324, 2020

    2020

  13. [13]

    Multi-resource fairness: Objectives, algorithms and performance,

    T. Bonald and J. Roberts, “Multi-resource fairness: Objectives, algorithms and performance,” in Proceedings of the 2015 ACM SIGMETRICS International Conference on Measurement and Modeling of Computer Systems , ser. SIGMETRICS ’15. New York, NY , USA: Association for Computing Machinery, 2015, p. 31–42. [Online]. Available: https://doi.org/10.1145/2745844.2745869

    work page doi:10.1145/2745844.2745869 2015

  14. [14]

    Hug: multi-resource fairness for correlated and elastic demands,

    M. Chowdhury, Z. Liu, A. Ghodsi, and I. Stoica, “Hug: multi-resource fairness for correlated and elastic demands,” in Proceedings of the 13th Usenix Conference on Networked Systems Design and Implementation , ser. NSDI’16. USA: USENIX Association, 2016, p. 407–424

    2016

  15. [15]

    A survey of the state-of-the-art in fair multi-resource allocations for data centers,

    P. Poullie, T. Bocek, and B. Stiller, “A survey of the state-of-the-art in fair multi-resource allocations for data centers,” IEEE Transactions on Network and Service Management , vol. 15, no. 1, pp. 169–183, 2018

    2018

  16. [16]

    Under- standing o-ran: Architecture, interfaces, algorithms, security, and re- search challenges,

    M. Polese, L. Bonati, S. D’oro, S. Basagni, and T. Melodia, “Under- standing o-ran: Architecture, interfaces, algorithms, security, and re- search challenges,” IEEE Communications Surveys & Tutorials, vol. 25, no. 2, pp. 1376–1411, 2023

    2023

  17. [17]

    Modelling of computational resources for 5g ran,

    S. Khatibi, K. Shah, and M. Roshdi, “Modelling of computational resources for 5g ran,” in 2018 European Conference on Networks and Communications (EuCNC). IEEE, 2018, pp. 1–5

    2018

  18. [18]

    On flexible association and placement in disaggregated ran designs,

    H. Hojeij, G. I. Ricardo, M. Sharara, S. Hoteit, V . V `eque, and S. Secci, “On flexible association and placement in disaggregated ran designs,” Computer Communications, vol. 238, p. 108166, 2025

    2025

  19. [19]

    Can fine-grained functional split benefit to the converged optical-wireless access networks in 5g and beyond?

    Y . Xiao, J. Zhang, and Y . Ji, “Can fine-grained functional split benefit to the converged optical-wireless access networks in 5g and beyond?” IEEE Transactions on Network and Service Management, vol. 17, no. 3, pp. 1774–1787, 2020

    2020

  20. [20]

    TS 38.214, 2023, release 17, v17.6.0

    TS 38.214: NR; Physical layer procedures for data , 3rd Generation Partnership Project (3GPP) Std. TS 38.214, 2023, release 17, v17.6.0. [Online]. Available: https://www.3gpp.org/ftp/Specs/archive/38 series/ 38.214/

    2023

  21. [21]

    A. W. Marshall, I. Olkin, and B. C. Arnold, Inequalities: theory of majorization and its applications . Springer, 1979

    1979

  22. [22]

    Delay scheduling: a simple technique for achieving locality and fairness in cluster scheduling,

    M. Zaharia, D. Borthakur, J. Sen Sarma, K. Elmeleegy, S. Shenker, and I. Stoica, “Delay scheduling: a simple technique for achieving locality and fairness in cluster scheduling,” in Proceedings of the 5th European conference on Computer systems , 2010, pp. 265–278

    2010

  23. [23]

    Faircloud: Sharing the network in cloud computing,

    L. Popa, G. Kumar, M. Chowdhury, A. Krishnamurthy, S. Ratnasamy, and I. Stoica, “Faircloud: Sharing the network in cloud computing,” in Proceedings of the ACM SIGCOMM 2012 conference on Applications, technologies, architectures, and protocols for computer communication , 2012, pp. 187–198

    2012

  24. [24]

    A theory of justice, harvard,

    J. Rawls, “A theory of justice, harvard,” Press, Cambridge, 1971

    1971

  25. [25]

    Fair optimization and networks: A survey,

    W. Ogryczak, H. Luss, M. Pi ´oro, D. Nace, and A. Tomaszewski, “Fair optimization and networks: A survey,” Journal of Applied Mathematics , vol. 2014, no. 1, p. 612018, 2014

    2014

  26. [26]

    Solving Max-Min fair resource allocations quickly on large graphs,

    P. Namyar, B. Arzani, S. Kandula, S. Segarra, D. Crankshaw, U. Krishnaswamy, R. Govindan, and H. Raj, “Solving Max-Min fair resource allocations quickly on large graphs,” in 21st USENIX Symposium on Networked Systems Design and Implementation (NSDI 24). Santa Clara, CA: USENIX Association, Apr. 2024, pp. 1937–

    2024

  27. [27]

    Available: https://www.usenix.org/conference/nsdi24/ presentation/namyar-solving

    [Online]. Available: https://www.usenix.org/conference/nsdi24/ presentation/namyar-solving

  28. [28]

    Axiomatic cost and surplus sharing,

    H. Moulin, “Axiomatic cost and surplus sharing,” Handbook of social choice and welfare , vol. 1, pp. 289–357, 2002

    2002

  29. [29]

    Fair resource allocation in systems with complete information sharing,

    F. Fossati, S. Hoteit, S. Moretti, and S. Secci, “Fair resource allocation in systems with complete information sharing,” IEEE/ACM Transactions on Networking, vol. 26, no. 6, pp. 2801–2814, 2018

    2018

  30. [30]

    Beyond dominant resource fairness: Extensions, limitations, and indivisibilities,

    D. C. Parkes, A. D. Procaccia, and N. Shah, “Beyond dominant resource fairness: Extensions, limitations, and indivisibilities,” ACM Transactions on Economics and Computation (TEAC) , vol. 3, no. 1, pp. 1–22, 2015

    2015

  31. [31]

    Dominant resource fairness in cloud computing systems with heterogeneous servers,

    W. Wang, B. Li, and B. Liang, “Dominant resource fairness in cloud computing systems with heterogeneous servers,” in IEEE INFOCOM 14 IEEE TRANSACTIONS ON NETWORKING, VOL. XX, YYYY 2014-IEEE Conference on Computer Communications . IEEE, 2014, pp. 583–591

    2014

  32. [32]

    {HUG}:{Multi- Resource} fairness for correlated and elastic demands,

    M. Chowdhury, Z. Liu, A. Ghodsi, and I. Stoica, “ {HUG}:{Multi- Resource} fairness for correlated and elastic demands,” in 13th USENIX symposium on networked systems design and implementation (NSDI 16), 2016, pp. 407–424

    2016

  33. [33]

    Multi-resource fairness: Objectives, algorithms and performance,

    T. Bonald and J. Roberts, “Multi-resource fairness: Objectives, algorithms and performance,” SIGMETRICS Perform. Eval. Rev. , vol. 43, no. 1, p. 31–42, Jun. 2015. [Online]. Available: https: //doi.org/10.1145/2796314.2745869

    work page doi:10.1145/2796314.2745869 2015

  34. [34]

    The measurement of the inequality of incomes,

    H. Dalton, “The measurement of the inequality of incomes,” The Economic Journal, vol. 30, no. 119, pp. 348–361, 1920

    1920

  35. [35]

    Disciplined convex- concave programming,

    X. Shen, S. Diamond, Y . Gu, and S. Boyd, “Disciplined convex- concave programming,” 2016 IEEE 55th Conference on Decision and Control (CDC) , pp. 1009–1014, 2016. [Online]. Available: https://stanford.edu/∼boyd/papers/dccp.html

    2016

  36. [36]

    Ec2instances.info - easy amazon ec2 instance comparison,

    Vantage, “Ec2instances.info - easy amazon ec2 instance comparison,” https://instances.vantage.sh/

  37. [37]

    TS 38.104: NR; Base Station (BS) radio transmission and reception,

    3GPP, “TS 38.104: NR; Base Station (BS) radio transmission and reception,” 3rd Generation Partnership Project (3GPP), Tech. Rep., 2023, release 17, v17.9.0. [Online]. Available: https://www.etsi.org/deliver/ etsi ts/138100 138199/138104/17.09.00 60/ts 138104v170900p.pdf

    2023

  38. [38]

    CVXPY: A Python-embedded modeling lan- guage for convex optimization,

    S. Diamond and S. Boyd, “CVXPY: A Python-embedded modeling lan- guage for convex optimization,” Journal of Machine Learning Research, vol. 17, no. 83, pp. 1–5, 2016

    2016

  39. [39]

    Disciplined Convex-Concave Programming

    X. Shen, S. Diamond, Y . Gu, and S. Boyd, “Disciplined convex-concave programming,” arXiv preprint arXiv:1604.02639 , 2016

    work page internal anchor Pith review Pith/arXiv arXiv 2016

  40. [40]

    A quantitative measure of fairness and discrimination,

    R. K. Jain, D.-M. W. Chiu, W. R. Hawe et al., “A quantitative measure of fairness and discrimination,” Eastern Research Laboratory, Digital Equipment Corporation, Hudson, MA , vol. 21, no. 1, pp. 2022–2023, 1984

    2022

  41. [41]

    Cost-efficient slicing in virtual radio access networks,

    S. Pramanik, A. Ksentini, and C. F. Chiasserini, “Cost-efficient slicing in virtual radio access networks,” Computer Communications, vol. 209, pp. 349–358, 2023. APPENDIX A ASSUMPTIONS ON DEPENDENCIES FOR SATURATION We now specify the assumptions on F that guarantee the Pareto efficiency of (DDRF) and (D-Util). Consider the dependency-aware multi-resource...

    2023