Comparison

DispatchAtlas is a toolkit for scheduling and optimization across distributed-computing environments, built around a curated, citation-backed catalog of problem instances — not only a solver library, a benchmark list, or a network simulator.

What sets DispatchAtlas apart

No single peer below is the differentiator — the combination is. DispatchAtlas pairs a curated, citation-backed catalog of scheduling and edge–fog–cloud optimization instances with a capability-typed solver registry, built-in non-parametric statistical comparison, and a public inspection portal whose in-browser suite builder composes ready-made or fully custom benchmark suites from the whole catalog and downloads them as JSON, CSV, or a bundle with no backend. Each of those pieces exists elsewhere; none of the peers below ships all of them together.

The nearest challengers deserve honest naming. jMetal ships an experimental study harness with non-parametric statistics, but no scheduling-instance catalog or inspection portal. job-shop-lib and the TU/e Job Shop Scheduling Benchmark ship rich scheduling instances, but no citation-per-instance curation, bundled statistics, or portal. WfCommons ships trace-backed workflow generators, but scores no optimizers and runs no statistics. DispatchAtlas is the project that holds all five capabilities at once.

This page is the canonical statement of that positioning; other surfaces link here rather than restating it.

How DispatchAtlas compares to related platforms

These tables compare platform capabilities — what each project ships — not performance. Performance evidence comes from the statistical analysis exports, never from this page. The tables are grouped into bands of comparable projects, keeping the same five capability columns throughout. Each peer links to its canonical citation; full sources are listed at the foot of the page. ✅ provided, ⚠️ adjacent or partial, ❌ not provided.

DispatchAtlas appears once, in the first band, and its ✅✅✅✅✅ row holds across every band that follows.

Scheduling benchmark platforms and solvers

OR-Tools CP-SAT is the dominant exact scheduling solver — DispatchAtlas does not compete with it. Optional exact adapters wrap solvers like it behind extras, and the capability matrix records exact optimization as an adapter capability: wrapped, not rivaled. job-shop-lib bundles classic instances alongside RL/GNN environments; the TU/e suite unifies JSP, FJSP, and online-FJSP environments for learning and non-learning methods; PyJobShop provides constraint-programming modeling evaluated on 9,000+ literature instances across machine and project scheduling; schlably is a deep-reinforcement-learning scheduling-experiment framework. None curates citation-per-instance provenance, ships statistical gates, or publishes an inspection portal — the ⚠️ marks flag instances that ship without curated citation metadata.

General optimization benchmarking and metaheuristic frameworks

COCO benchmarks a different problem class (continuous black-box functions); its suite family spans single-, bi-objective, large-scale, mixed-integer, and constrained variants with a public data archive. IOHprofiler benchmarks iterative heuristics and offers a hosted web analyzer. jMetal and jMetalPy are included precisely because they are not strawmen: they ship experimental studies with non-parametric statistics, yet no scheduling catalog or portal. Nevergrad provides rank plots and an experiment-and-plotting module over hundreds of optimizers, but not non-parametric inference. pymoo, mealpy, and NiaPy provide rich algorithm collections and synthetic test functions rather than a capability-typed registry oriented to scheduling.

Fog and edge simulators

These model edge–fog–cloud networks as discrete-event simulations (⚠️), not as a benchmark suite an optimizer is scored on. EdgeCloudSim v5.0 runs on the current CloudSim 7.0.0-alpha core with new tutorials. YAFS is the Python fog simulator closest in spirit to the continuum families, and its placement problems are what DispatchAtlas re-expresses as scored benchmark instances; EdgeSimPy is the most recent Python edge-resource-management simulator.

Cloud and workflow simulation platforms

These are simulators and trace/generator frameworks, not optimizer benchmarks — the same honest framing the fog band uses. WfCommons's trace-to-generator-to-benchmark pipeline is the strongest conceptual analogue of the platform's trace-backed ambitions, and it is exactly where DispatchAtlas is weakest today: one calibrated-synthetic profile, the ⚠️ trace-backed row in the capability matrix below.

Optuna is complementary tooling, not a peer: it is a hyperparameter tuner, so solver hyperparameters can be tuned by it, but it ships no scheduling instances and no provenance story — which is why "why not just use Optuna?" has a one-line answer. Cite Optuna (Akiba et al., KDD 2019).

How to read the marks:

• Curated scheduling-instance catalog — a versioned, citation-backed set of scheduling and permutation problem instances. The scheduling peers ship instances but not citation-per-instance curation; the optimization frameworks ship synthetic test problems; the simulators ship no optimization-instance catalog.
• Edge–fog–cloud as benchmark instances — heterogeneous placement and communication problems exposed as optimization instances. The simulators model edge–fog–cloud networks (⚠️), but as discrete-event simulations, not as a benchmark suite an optimizer is scored on.
• Capability-typed solver registry — solvers carry declared objectives, constraints, and capability metadata. The metaheuristic libraries provide rich algorithm collections (⚠️) but not a capability-typed registry oriented to scheduling.
• Built-in statistical comparison — non-parametric tests and effect sizes bundled with the toolkit. COCO, IOHprofiler, and jMetal ship dedicated statistical tooling.
• Public inspection portal — a browsable, public-safe data surface. COCO publishes a workshop data archive (⚠️); IOHanalyzer offers a hosted web analyzer.

What DispatchAtlas does not provide

This is the honest reverse gap — each item names the peer that has what DispatchAtlas does not, and mirrors an existing ❌ or ⚠️ row in the capability matrix below.

• No discrete-event simulation — the fog and cloud simulators (CloudSim, iFogSim2, EdgeCloudSim, WRENCH) model time-stepped network and platform dynamics; DispatchAtlas scores optimizers on instances instead.
• No dynamic-arrival or online-rescheduling generators — the TU/e suite ships online-FJSP environments; every bundled DispatchAtlas generator is static.
• No learning-agent training environments — job-shop-lib and schlably ship RL/GNN environments; DispatchAtlas's learning interfaces are partial.
• Single-host campaign execution — Nevergrad executes large parallel experiment batteries; DispatchAtlas runs sequential, threaded, bounded, and replay execution on one host.
• One calibrated-synthetic trace profile — WfCommons ships validated trace-derived generators over real workflow archives; DispatchAtlas bundles a single calibrated-synthetic profile.

Sources

• COCO — Hansen, Auger, Ros, Mersmann, Tušar, Brockhoff. COCO: a platform for comparing continuous optimizers in a black-box setting. Optimization Methods and Software 36(1), 2021. https://doi.org/10.1080/10556788.2020.1808977
• IOHprofiler — de Nobel, Ye, Vermetten, Wang, Doerr, Bäck. IOHexperimenter: Benchmarking Platform for Iterative Optimization Heuristics. Evolutionary Computation 32(3):205–210, 2024. https://doi.org/10.1162/evco_a_00342 · IOHanalyzer — https://doi.org/10.1145/3510426
• pymoo — Blank, Deb. pymoo: Multi-Objective Optimization in Python. IEEE Access 8, 2020. https://doi.org/10.1109/ACCESS.2020.2990567
• mealpy — Van Thieu, Mirjalili. MEALPY: An open-source library for latest meta-heuristic algorithms in Python. Journal of Systems Architecture 139, 2023. https://doi.org/10.1016/j.sysarc.2023.102871
• NiaPy — Vrbančič, Brezočnik, Mlakar, Fister, Fister Jr. NiaPy: Python microframework for building nature-inspired algorithms. Journal of Open Source Software 3(23), 2018. https://doi.org/10.21105/joss.00613
• iFogSim2 — Mahmud, Pallewatta, Goudarzi, Buyya. iFogSim2: An extended iFogSim simulator … Journal of Systems and Software 190, 2022. https://doi.org/10.1016/j.jss.2022.111351
• EdgeCloudSim — Sonmez, Ozgovde, Ersoy. EdgeCloudSim: An environment for performance evaluation of edge computing systems. Transactions on Emerging Telecommunications Technologies 29(11), 2018. https://doi.org/10.1002/ett.3493
• LEAF — Wiesner, Thamsen. LEAF: Simulating Large Energy-Aware Fog Computing Environments. IEEE ICFEC, 2021. https://doi.org/10.1109/ICFEC51620.2021.00012
• job-shop-lib — Ariño Fernández. job-shop-lib. arXiv:2506.13781, 2025. https://arxiv.org/abs/2506.13781
• Job Shop Scheduling Benchmark — Reijnen, van Straaten, Bukhsh, Zhang. arXiv:2308.12794, 2023. https://arxiv.org/abs/2308.12794
• PyJobShop — Lan, Berkhout. PyJobShop. arXiv:2502.13483, 2025. https://arxiv.org/abs/2502.13483
• schlably — Waubert de Puiseau et al. schlably. SoftwareX 22:101383, 2023. https://doi.org/10.1016/j.softx.2023.101383
• OR-Tools CP-SAT — Perron, Didier, Gay. CP-SAT. CP 2023. https://doi.org/10.4230/LIPIcs.CP.2023.3
• jMetal — Durillo, Nebro. jMetal. Advances in Engineering Software 42(10), 2011. https://doi.org/10.1016/j.advengsoft.2011.05.014 · jMetalPy — Benítez-Hidalgo et al. Swarm and Evolutionary Computation 51:100598, 2019. https://doi.org/10.1016/j.swevo.2019.100598
• Nevergrad — Rapin, Teytaud. Nevergrad. 2018 (software citation). https://github.com/facebookresearch/nevergrad
• YAFS — Lera, Guerrero, Juiz. YAFS. IEEE Access 7, 2019. https://doi.org/10.1109/ACCESS.2019.2927895
• EdgeSimPy — Souza et al. EdgeSimPy. Future Generation Computer Systems 148, 2023. https://doi.org/10.1016/j.future.2023.06.013
• CloudSim — Calheiros, Ranjan, Beloglazov, De Rose, Buyya. CloudSim. Software: Practice and Experience 41(1), 2011. https://doi.org/10.1002/spe.995 · CloudSim 7G — Andreoli et al. Software: Practice and Experience 55(6), 2025. https://doi.org/10.1002/spe.3413
• WfCommons — Coleman et al. WfCommons. Future Generation Computer Systems 128, 2022. https://doi.org/10.1016/j.future.2021.09.043
• WRENCH — Casanova et al. WRENCH. Future Generation Computer Systems 112, 2020. https://doi.org/10.1016/j.future.2020.05.030
• Optuna — Akiba, Sano, Yanase, Ohta, Koyama. Optuna: A Next-generation Hyperparameter Optimization Framework. KDD 2019. https://doi.org/10.1145/3292500.3330701

DispatchAtlas stance

Capability Matrix

Each status is verified against the shipped packages and the public portal bundles named in the basis column: ✅ supported today, ⚠️ partial, ❌ absent.