pith. sign in

arxiv: 2604.19181 · v1 · submitted 2026-04-21 · 💻 cs.DC

YAIFS: Yet (not) Another Intelligent Fog Simulator: A Framework for Agent-Driven Computing Continuum Modeling & Simulation

Isaac Lera , Carlos Guerrero This is my paper

Pith reviewed 2026-05-10 02:20 UTC · model grok-4.3

classification 💻 cs.DC
keywords YAIFSfog simulatorModel Context Protocolinteractive simulationagent coordinationadaptive systemsedge computingAI-driven experimentation
0
0 comments X

The pith

YAIFS turns fog simulations into interactive environments where external agents observe, control, and adapt executions via a shared protocol.

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

The paper introduces YAIFS as an updated simulator that changes how researchers interact with fog computing models. Instead of static runs, it exposes the simulation through an API and service layer so outside entities can watch, steer, and change what happens during execution. The central addition is the Model Context Protocol, which gives different kinds of agents a common set of tools to read state, call actions, and work together. This matters because it removes the need to build separate connections for each new agent or experiment, making it simpler to test adaptive behaviors in dynamic edge systems.

Core claim

We present YAIFS, an evolution of earlier fog simulators, that redefines simulation as an interactive, service-oriented environment. A layered architecture exposes the simulation through a unified API and service interface, allowing external entities to observe, control, and modify execution. The central contribution is the integration of the Model Context Protocol as a standardized interaction layer. Through this protocol, heterogeneous agents access state, invoke actions, and coordinate behavior using a common set of tools. Two scenarios illustrate the approach: an LLM-based assistant that enables natural language control and a multi-agent setting where agents monitor conditions and adapt

What carries the argument

The Model Context Protocol (MCP), a standardized interaction layer that lets heterogeneous agents access simulation state, invoke actions, and coordinate using shared tools without custom per-simulation code.

Load-bearing premise

That adding the standardized protocol enables effective agent coordination and adaptation under changing conditions without performance penalties or compatibility problems.

What would settle it

Running the multi-agent adaptation scenario with rapidly shifting workloads and finding that agents fail to adjust placement decisions correctly or that overall simulation speed drops due to protocol overhead.

Figures

Figures reproduced from arXiv: 2604.19181 by Carlos Guerrero, Isaac Lera.

Figure 1
Figure 1. Figure 1: Layered visibility model 3 [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Example topology of a cloud–continuum infrastructure composed of 3 [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Evolution of agent strategies over time for each application [PITH_FULL_IMAGE:figures/full_fig_p020_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Evolution of multi-agent actions over time [PITH_FULL_IMAGE:figures/full_fig_p021_4.png] view at source ↗
read the original abstract

Simulation plays a key role in the design and evaluation of distributed systems, yet it is often treated as a static tool with limited interaction capabilities. In this work, we present Yet (not) Another Intelligent Fog Simulator (YAIFS), and evolution of YAFS that redefines simulation as an interactive, service-oriented environment. YAIFS introduce a layered architecture that exposes the simulation through a unified API and service interface, enabling external entities to observe, control, and modify its execution. A central contribution is the integration of the Model Context Protocol (MCP) as a standardized interaction layer between agents and the simulation. Through MCP, heterogeneous agents can access state, invoke actions and coordinate behavior using a common set of tools, decoupling agent experimentation workflows. We illustrate these capabilities through two scenarios: an LLM-based assistant that enable natural language control of simulations, and a multi-agent setting where agents monitor system conditions and adapt placement decisions at runtime. These scenarios demonstrate how MCP structures agent-simulation interaction and enable adaptive behavior under dynamic workloads. The proposed approach transforms simulation into an interactive and programmable environment, opening new directions for AI-driven experimentation in cloud-edge systems. The implementation is publicly available at: http://github.com/acsicuib/YAIFS

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 / 3 minor

Summary. The manuscript presents YAIFS as an evolution of the YAFS fog simulator, introducing a layered architecture that exposes the simulation via a unified API and service interface. A central contribution is the integration of the Model Context Protocol (MCP) as a standardized interaction layer, allowing heterogeneous agents to access state, invoke actions, and coordinate using common tools, thereby decoupling agent experimentation workflows. This is illustrated through two scenarios: an LLM-based assistant enabling natural language control of simulations and a multi-agent setting where agents monitor conditions and adapt placement decisions at runtime under dynamic workloads. The implementation is publicly available on GitHub.

Significance. If the effectiveness claims hold, YAIFS could advance AI-driven experimentation in cloud-edge systems by converting simulations into interactive, programmable environments. The public code release supports reproducibility and community extension, which is a clear strength for a systems framework paper.

major comments (2)
  1. [§5] §5 (Illustrative Scenarios): The two scenarios are presented purely at a descriptive level with no quantitative metrics (e.g., interaction latency, throughput, coordination success rate, or overhead versus direct API access), no ablation studies, and no workload scaling results. This directly undermines the central claim that MCP enables effective adaptive behavior under dynamic workloads via seamless integration without performance penalties or compatibility barriers.
  2. [§3] §3 (Architecture and MCP Integration): The assertion that MCP 'decouples agent experimentation workflows' and structures interactions without compatibility barriers is stated without any compatibility tests across agent types or empirical comparison to non-MCP interfaces, leaving the decoupling benefit as an unverified assumption rather than demonstrated outcome.
minor comments (3)
  1. [Abstract] Abstract: 'YAIFS introduce a layered architecture' contains a subject-verb agreement error and should read 'YAIFS introduces a layered architecture'.
  2. [Abstract] Abstract: 'an LLM-based assistant that enable natural language control' should be corrected to 'an LLM-based assistant that enables natural language control'.
  3. [§3] The manuscript would benefit from a dedicated subsection or table summarizing the MCP tool set (state access, actions, coordination primitives) to improve clarity of the standardized interface.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and detailed feedback on our manuscript. We have carefully reviewed the major comments and provide point-by-point responses below. Where the comments identify gaps in empirical support, we agree and outline specific revisions to address them.

read point-by-point responses

  1. Referee: [§5] §5 (Illustrative Scenarios): The two scenarios are presented purely at a descriptive level with no quantitative metrics (e.g., interaction latency, throughput, coordination success rate, or overhead versus direct API access), no ablation studies, and no workload scaling results. This directly undermines the central claim that MCP enables effective adaptive behavior under dynamic workloads via seamless integration without performance penalties or compatibility barriers.

    Authors: We acknowledge that Section 5 presents the scenarios at an illustrative level without accompanying quantitative metrics, ablation studies, or scaling results. This is a valid observation, as the primary contribution of the work is the layered architecture and MCP integration rather than a performance benchmark study. To strengthen the claims regarding seamless integration and adaptive behavior, the revised manuscript will expand Section 5 to include basic quantitative measurements: interaction latency and overhead for MCP versus direct API access, coordination success rates in the multi-agent scenario, and results under varying dynamic workloads. These additions will be presented as supporting evidence for the framework's effectiveness. revision: yes

  2. Referee: [§3] §3 (Architecture and MCP Integration): The assertion that MCP 'decouples agent experimentation workflows' and structures interactions without compatibility barriers is stated without any compatibility tests across agent types or empirical comparison to non-MCP interfaces, leaving the decoupling benefit as an unverified assumption rather than demonstrated outcome.

    Authors: The decoupling benefit follows from the design of MCP as a standardized protocol that exposes a uniform set of tools and state access methods, allowing any compliant agent to interact without custom per-agent code. The two scenarios in the manuscript already demonstrate this with an LLM-based agent and a separate multi-agent system using the identical MCP layer. However, we agree that explicit compatibility tests and comparisons to non-MCP interfaces would make the claim more robust. In the revision, we will add a short subsection or paragraph in Section 3 that discusses compatibility by design and includes a simple empirical comparison of interface uniformity and setup effort across agent types. revision: yes

Circularity Check

0 steps flagged

No circularity in framework description

full rationale

The paper presents YAIFS as an evolution of the prior YAFS simulator, with a layered architecture and integration of the Model Context Protocol (MCP) for agent-simulation interaction. It describes the design, API exposure, and two high-level illustrative scenarios (LLM assistant and multi-agent placement) without any equations, fitted parameters, quantitative predictions, or derivation steps. Claims about decoupling workflows and enabling adaptive behavior are design assertions supported by scenario descriptions rather than results that reduce to inputs by construction. No self-citation load-bearing steps, uniqueness theorems, or ansatzes appear in the provided text to create circularity. This is a standard software framework contribution whose central claims do not rely on self-referential reductions.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 1 invented entities

The central claims rest on the introduction of MCP and the layered service interface together with the domain assumption that agents can usefully coordinate via the provided tools.

axioms (1)
  • domain assumption Heterogeneous agents can effectively coordinate behavior and adapt decisions using a common standardized tool set provided by MCP
    Invoked to support the multi-agent runtime adaptation scenario.
invented entities (1)
  • Model Context Protocol (MCP) no independent evidence
    purpose: Standardized interaction layer enabling agents to access state, invoke actions, and coordinate with the simulator
    Presented as a central new contribution without external references or independent validation data.

pith-pipeline@v0.9.0 · 5521 in / 1200 out tokens · 45218 ms · 2026-05-10T02:20:47.657224+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

30 extracted references · 30 canonical work pages · 1 internal anchor

  1. [1]

    Simulation Modelling Practice and Theory , volume=

    Quality matters: A comprehensive comparative study of edge computing simulators , author=. Simulation Modelling Practice and Theory , volume=. 2025 , publisher=

    work page 2025

  2. [2]

    A survey and taxonomy of simulation environments modelling fog computing , journal =

    Andras Markus and Attila Kertesz , keywords =. A survey and taxonomy of simulation environments modelling fog computing , journal =. 2020 , note =. doi:https://doi.org/10.1016/j.simpat.2019.102042 , url =

    work page doi:10.1016/j.simpat.2019.102042 2020

  3. [3]

    and hashem, S.H

    Hussein, B.A. and hashem, S.H. , year =. iFogger: A New Framework to Simulate Fog Computing Resource Placement inside OMNeT++ Environment , volume =. Mustansiriyah Journal of Pure and Applied Sciences , doi =

    work page

  4. [4]

    2020 12th International Conference on Communication Software and Networks (ICCSN) , pages=

    SatEdgeSim: A toolkit for modeling and simulation of performance evaluation in satellite edge computing environments , author=. 2020 12th International Conference on Communication Software and Networks (ICCSN) , pages=. 2020 , organization=

    work page 2020

  5. [5]

    doi:10.5281/zenodo.10953027 , url =

    Del Pozo Puñal, Elías and Garcia-Carballeira, Felix , title =. doi:10.5281/zenodo.10953027 , url =

    work page doi:10.5281/zenodo.10953027

  6. [6]

    EdgeAISim: A toolkit for simulation and modelling of AI models in edge computing environments , journal =

    Aadharsh Roshan Nandhakumar and Ayush Baranwal and Priyanshukumar Choudhary and Muhammed Golec and Sukhpal Singh Gill , keywords =. EdgeAISim: A toolkit for simulation and modelling of AI models in edge computing environments , journal =. 2024 , issn =. doi:https://doi.org/10.1016/j.measen.2023.100939 , url =

    work page doi:10.1016/j.measen.2023.100939 2024

  7. [7]

    Computer Science and Information Systems , volume=

    PureEdgeSim: A simulation framework for performance evaluation of cloud, edge and mist computing environments , author=. Computer Science and Information Systems , volume=

    work page

  8. [8]

    Software: Practice and Experience , volume=

    iFogSim: A toolkit for modeling and simulation of resource management techniques in the Internet of Things, Edge and Fog computing environments , author=. Software: Practice and Experience , volume=. 2017 , publisher=

    work page 2017

  9. [9]

    ECLYPSE: A Python Framework for Simulation and Emulation of the Cloud‐Edge Continuum , volume=

    Massa, Jacopo and De Caro, Valerio and Forti, Stefano and Dazzi, Patrizio and Bacciu, Davide and Brogi, Antonio , year=. ECLYPSE: A Python Framework for Simulation and Emulation of the Cloud‐Edge Continuum , volume=. Journal of Software: Evolution and Process , publisher=. doi:10.1002/smr.70081 , number=

    work page doi:10.1002/smr.70081

  10. [10]

    Applied Sciences , VOLUME =

    Markus, Andras and Biro, Mate and Skala, Karolj and Šojat, Zorislav and Kertesz, Attila , TITLE =. Applied Sciences , VOLUME =. 2022 , NUMBER =

    work page 2022

  11. [11]

    IEEE Access , title=

    I. IEEE Access , title=. 2019 , volume=. doi:10.1109/ACCESS.2019.2927895 , ISSN=

    work page doi:10.1109/access.2019.2927895 2019

  12. [12]

    IEEE Transactions on Network and Service Management , year=

    Multi-Objective Deep Reinforcement Learning Assisted Service Function Chains Placement , author=. IEEE Transactions on Network and Service Management , year=

    work page

  13. [13]

    Brogi, S

    Brogi, Antonio and Forti, Stefano and Guerrero, Carlos and Lera, Isaac , title =. Software: Practice and Experience , volume =. doi:https://doi.org/10.1002/spe.2766 , url =. https://onlinelibrary.wiley.com/doi/pdf/10.1002/spe.2766 , abstract =

    work page doi:10.1002/spe.2766

  14. [14]

    A survey on service function chaining , journal =

    Deval Bhamare and Raj Jain and Mohammed Samaka and Aiman Erbad , keywords =. A survey on service function chaining , journal =. 2016 , issn =. doi:https://doi.org/10.1016/j.jnca.2016.09.001 , url =

    work page doi:10.1016/j.jnca.2016.09.001 2016

  15. [15]

    Multi-objective application placement in fog computing using graph neural network-based reinforcement learning

    Isaac Lera and Carlos Guerrero , title =. The Journal of Supercomputing , year =. doi:10.1007/s11227-024-06439-5 , url =

    work page doi:10.1007/s11227-024-06439-5

  16. [16]

    Proximal Policy Optimization Algorithms

    Proximal policy optimization algorithms , author=. arXiv preprint arXiv:1707.06347 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  17. [17]

    European journal of operational research , volume=

    Topsis for MODM , author=. European journal of operational research , volume=. 1994 , publisher=

    work page 1994

  18. [18]

    Expert Systems with applications , volume=

    A state-of the-art survey of TOPSIS applications , author=. Expert Systems with applications , volume=. 2012 , publisher=

    work page 2012

  19. [19]

    Soft Computing , year =

    Vinay Pandey and Komal and Hasan Dincer , title =. Soft Computing , year =. doi:10.1007/s00500-023-09011-0 , url =

    work page doi:10.1007/s00500-023-09011-0

  20. [20]

    Baranwal, Gaurav and Yadav, Ravi and Vidyarthi, Deo Prakash , title =. Mob. Netw. Appl. , month = oct, pages =. 2020 , issue_date =. doi:10.1007/s11036-020-01563-x , abstract =

    work page doi:10.1007/s11036-020-01563-x 2020

  21. [21]

    Analyzing the Applicability of a Multi-Criteria Decision Method in Fog Computing Placement Problem , year=

    Lera, Isaac and Guerrero, Carlos and Juiz, Carlos , booktitle=. Analyzing the Applicability of a Multi-Criteria Decision Method in Fog Computing Placement Problem , year=

    work page

  22. [22]

    Applied Sciences ,year=

    A Fuzzy-Based Method for Objects Selection in Blockchain-Enabled Edge-IoT Platforms Using a Hybrid Multi-Criteria Decision-Making Model ,author=. Applied Sciences ,year=

    work page

  23. [23]

    Cluster Computing ,year=

    Blockchain-driven optimization of IoT in mobile edge computing environment with deep reinforcement learning and multi-criteria decision-making techniques ,author=. Cluster Computing ,year=

    work page

  24. [24]

    and Greco, S

    Figueira, J. and Greco, S. and Ehrgott, M. , biburl =

    work page

  25. [25]

    ACM Transactions on Sensor Networks ,year=

    UETOPSIS: A Data-Driven Intelligence Approach to Security Decisions for Edge Computing in Smart Cities ,author=. ACM Transactions on Sensor Networks ,year=

    work page

  26. [26]

    Trust-Oriented IoT Service Placement for Smart Cities in Edge Computing , year=

    Xu, Xiaolong and Liu, Xihua and Xu, Zhanyang and Dai, Fei and Zhang, Xuyun and Qi, Lianyong , journal=. Trust-Oriented IoT Service Placement for Smart Cities in Edge Computing , year=

    work page

  27. [27]

    Transactions on Emerging Telecommunications Technologies ,year=

    Edge server placement for offloading real‐time statistics tasks of bus passenger ,author=. Transactions on Emerging Telecommunications Technologies ,year=

    work page

  28. [28]

    Computing ,year=

    Enhancing task scheduling and QoS optimization in mobile edge computing via microservice-oriented container selection ,author=. Computing ,year=

    work page

  29. [29]

    QoS-Aware Deployment of IoT Applications Through the Fog , year=

    Brogi, Antonio and Forti, Stefano , journal=. QoS-Aware Deployment of IoT Applications Through the Fog , year=

    work page

  30. [30]

    ACM Comput

    Taleb, Imane and Guillaume, Jean-Loup and Duthil, Benjamin , title =. ACM Comput. Surv. , month = jun, articleno =. 2025 , issue_date =. doi:10.1145/3729214 , abstract =

    work page doi:10.1145/3729214 2025