{"id":"8673b325-2bb9-4972-b3dd-2802d58369e0","arxiv_id":"2502.09783","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The authors propose SLICES, a pan-European research infrastructure to make networking experiments more credible and reproducible by adapting the ESFRI scientific-instrument model.","lead":"This paper presents SLICES, a proposed European infrastructure to serve as a shared, large-scale scientific testbed for computer networking research, with tools for managing data and reproducing experiments. It argues that networking needs the kind of centralized instruments that physics and biology use.","discovery_kind":"review","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The paper's own introduction lists rapid technology change and community fragmentation as reasons a networking research instrument is impossible, but the rest of the paper never addresses those obstacles; the central feasibility claim is therefore unsupported.","rationale":"The reader's weakest assumption identifies the same load-bearing concern: a slow-moving ESFRI facility can remain relevant only if the field's technology and community fragmentation do not outrun its governance. The paper itself places this objection at the front door in Section I, then never returns to it. That is an internal gap, not merely a disagreement with an external consensus. The paper is a position and design paper without experimental or quantitative claims beyond planning estimates, so the main reviewable question is whether the argument is coherent. The technical descriptions of 5G, O-RAN, SDN/NFV, and MANO align with the cited literature, and the paper honestly frames the architecture as preliminary. It also gives credit to relevant open-source efforts and cites a design-study deliverable. Those are real strengths. However, the central claim has two parts: that SLICES is needed and that it is feasible. The need part is supported mainly by assertion, which is acceptable in a position paper. The feasibility part is more load-bearing because the paper's own introduction raises strong reasons to doubt it: technology changes quickly, future questions are hard to predict, and the community is fragmented. The paper does not explain how a multi-year ESFRI process with distributed nodes in 15 countries will track a technology that changes on an 18-month 3GPP release cadence, nor how it will overcome the fragmentation that has historically led researchers to build in-house testbeds. If the SLICES design study includes a concrete technology-refresh mechanism aligned with that cadence, the concern would be mitigated. If not, the paper overclaims feasibility. The reader's CONDITIONAL verdict already captures this uncertainty, so I recommend no change to the verdict: the proposal is plausible and well-presented, but the load-bearing feasibility assumption needs evidence from the design-study documentation or from successor-experiment usage data.","tokens_in":20136,"tokens_out":4362,"duration_ms":50579,"concrete_test":"Check the SLICES Design Study deliverables (especially D4.2, cited in Section VII) and SLICES-SC public documents for a defined hardware/software refresh cycle and community-adoption governance. Then compare the refresh cadence with the 3GPP Release cadence (roughly 18 months) and O-RAN specification evolution. If no explicit refresh mechanism exists, or if the interval between design freeze and operational upgrade exceeds two 3GPP releases, the facility would run an obsolete technology baseline for most of its lifecycle, and the feasibility claim fails. A complementary check: measure post-funding usage trajectories of GENI and FIRE federated testbeds; a steep decline after initial projects would show that large instruments do not automatically retain fragmented communities.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section I enumerates three reasons why 'prototyping a scientific instrument' for networking is 'often considered as impossible or irrelevant': (i) unpredictable future landscape, (ii) technology evolving too quickly, and (iii) fragmented community. The paper then proceeds to propose SLICES without ever resolving these obstacles. The central claim that SLICES is needed and feasible depends on a slow-moving, multi-national ESFRI facility remaining scientifically relevant and attracting a fragmented community across a decade-scale lifecycle. The architecture sections (V-VII) are anchored to current technologies: 5G NR, O-RAN, NFV-MANO, Kubernetes. No technology-refresh governance, no replacement mechanism for obsolete RAN/Core components, and no evidence from predecessor facilities (GENI, FIRE, PAWR) that large instruments stayed relevant after their founding technology wave are provided. The planning figures in Section VI (5,000 users, 50GB/user on nodes, 1TB/user on cloud, 0.25-1PB + 5PB storage) are asserted without derivation, but the more fundamental gap is the missing response to the paper's own feasibility objections. Without an explicit refresh and adoption mechanism, the claim that SLICES will be the reference instrument for beyond-5G/6G research is not established.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper argues that computer-networking research needs a structured, experimentally driven methodology supported by a dedicated scientific instrument, and it presents the SLICES initiative as the European answer to that need. After introducing the ESFRI framework, the paper reviews enabling technologies (SDN/NFV, network slicing, disaggregation, CUPS, MEC), illustrates them with a 5G/O-RAN example, proposes a layered architecture for SLICES, and describes the research-lifecycle, FAIR-data, and EOSC-interoperability components of the design. The paper is a position/design document: it contains no experimental evaluation, and its contribution is a community-level proposal rather than a falsifiable technical result.","tokens_in":20325,"tokens_out":3475,"duration_ms":38984,"significance":"If realized as described, SLICES would be a substantial community resource: it would provide a pan-European, federated, programmable instrument for networking research, with explicit commitments to FAIR data, reproducible experiments, and EOSC interoperability. The paper's strengths are its broad and mostly accurate synthesis of current open-source RAN/core/MANO components, its concrete layered architecture (Section V), and its clear embedding of the proposal in the ESFRI lifecycle and Open Science agenda. The main weakness is that the feasibility claim rests on programmatic assertions: the paper acknowledges the standard objections to such an instrument but does not resolve them, and the sizing estimates and \"lessons learned\" framing are not supported by evidence.","major_comments":[{"comment":"The paper opens by listing three objections to a networking scientific instrument: (i) unpredictable future questions, (ii) rapid technology evolution, and (iii) community fragmentation. The remainder of the paper describes the proposed architecture and technology choices but never returns to these objections. The architecture in Section V and the technology survey in Section IV are anchored to 5G NR, O-RAN, NFV-MANO, and Kubernetes; no mechanism is specified for refreshing or retiring components as the technology wave moves on. Because the value of a decade-scale ESFRI facility depends on continued relevance and community adoption, the central feasibility claim requires an explicit treatment of technology-refresh governance, incentives for a fragmented community to converge on one instrument, and evidence from predecessor facilities (GENI, FIRE, PAWR) about how large testbeds fared through technology transitions.","section":"Section I and Sections V-VII"},{"comment":"The sentence \"Our preliminary estimations for SLICES include up to 5,000 users ... 0.25PB-1PB of data storage ... and 5PB for the cloud-based datacenter\" presents the sizing figures as facts, but no derivation, source, or sensitivity analysis is provided. These numbers are load-bearing for the ESFRI-scale feasibility argument: an instrument that is either grossly oversized or undersized would undermine the proposal. The authors should either derive the estimates from usage data of predecessor platforms (OneLab, PlanetLab Europe, GENI, PAWR) or explicitly label them as illustrative planning assumptions with stated caveats.","section":"Section VI"},{"comment":"The abstract claims that the paper \"reports lessons learned from the design and operation of test platforms,\" but the body is largely a survey of enabling technologies and a proposed architecture. There is no systematic discussion of what succeeded or failed in PlanetLab, ORBIT, GENI, or FIRE, no metrics of adoption, reproducibility, or sustainability, and no analysis of why SLICES would avoid the fragmentation that the paper attributes to current practice. If the \"lessons learned\" claim is retained, the relevant evidence and analysis need to be included; otherwise the claim should be softened to \"experience and design considerations.\"","section":"Abstract and Sections III-VII"}],"minor_comments":[{"comment":"The URLs in the srsLTE and SD-RAN rows appear to be swapped: the srsLTE row points to openairinterface.org and the SD-RAN row points to github.com/srsran/srsRAN.","section":"Table I"},{"comment":"The phrase \"the FlexRIC platform (also called as FlexRAN)\" conflates two related but distinct pieces of software; the relationship between FlexRIC and FlexRAN should be clarified.","section":"Section IV.A"},{"comment":"There are several typographical issues: \"lets consider\" in Section VIII, \"life-cyle\" in Section IX, and \"computer networks needs\" in the abstract.","section":"Section VIII and Section IX"},{"comment":"References [61] and [62] are bare URLs without titles or authors; they should be completed so that readers can verify the sources.","section":"References [61] and [62]"}],"recommendation":"major_revision","confidential_remarks":"This is a position/design paper rather than a technical contribution, and the referee report applies the appropriate standard: the central claim is judged on coherence and support rather than on experimental validation. The main risk is that the proposal's feasibility rests on the unresolved objections listed in its own introduction; adding an explicit governance/refresh/adoption discussion and substantiating or re-labeling the sizing estimates would materially strengthen the paper. The heavy citation of the authors' own projects (OAI, FlexRIC, OneLab, SLICES) is understandable in this context but should be balanced by independent evidence from the predecessor facilities."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague, this is a position paper, not a research result, and it should be read as one. The genuinely new thing is the institutional packaging: taking the ESFRI scientific-instrument lifecycle and applying it to a federated networking testbed, with FAIR/EOSC interoperability as a first-class requirement. The paper also does a decent job of summarizing the current open-source 5G/O-RAN/NFV-MANO landscape, and the layered architecture in Figure 7 is a reasonable synthesis of where testbed federation has been going since SFA. It is clearly written and honest about being preliminary.\n\nThe soft spots are real, though. The introduction lists three reasons people think a networking scientific instrument is impossible — unpredictable landscape, fast-moving technology, fragmented community — and then the paper never returns to them. The architecture is anchored to 5G NR, O-RAN, Kubernetes, and NFV-MANO as they exist today; there is no technology-refresh governance, no discussion of how the facility stays relevant on a decade-scale ESFRI lifecycle, and no evidence from GENI/FIRE/PAWR that large instruments survived their founding technology wave. That tension sits right at the center of the feasibility claim, so it cannot be waved off as a minor omission. The planning numbers (5,000 users, 50 GB/user on nodes, 1 TB/user on cloud, 0.25–1 PB plus 5 PB) are asserted without derivation; they are rough sizing estimates, fine for a proposal, but they are not analysis. And the paper repeatedly defers specifics to the authors' own design-study deliverables, which limits how much a reader can evaluate on the page.\n\nThe citation pattern is mostly appropriate — the technical descriptions are backed by the standards and open-source project references you would expect. There is a clear institutional bias toward the authors' own tools (FlexRIC, OneLab, SLICES design studies), but that is not a logical flaw; it is just something to keep in mind.\n\nWho is this for? People who want a single entry point to what SLICES is, what it is trying to build, and how it relates to EOSC and FAIR. It is not a paper that settles whether the facility will work. As a position paper it deserves serious refereeing — I would send it to review. My verdict is conditional: the paper succeeds as a statement of intent and a technical overview, but the answer to its own feasibility objections is missing, and that is exactly the part that needs outside scrutiny.","headline":"A clear, honest position paper for SLICES, but it raises its own feasibility objections and never answers them.","tokens_in":20981,"tokens_out":2276,"would_cite":true,"duration_ms":21692,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Computer networking research should be conducted on a shared, calibrated scientific instrument, and SLICES is that instrument.","keywords":["experimentally-driven networking research","scientific instrument","testbed federation","reproducibility","FAIR data","5G/6G networks","SDN/NFV","research lifecycle"],"falsifier":"See whether the fraction of networking papers whose experiments can be independently reproduced on SLICES rises above today's baseline within five years of the facility opening; if it does not, the paper's central claim is not borne out.","tokens_in":19887,"feed_emoji":"🔬","tokens_out":4943,"duration_ms":47718,"temperature":0.7,"pith_summary":"The paper argues that computer networking research has matured to the point where it needs a formal scientific instrument, not just scattered in-house testbeds, and that the European SLICES initiative is that instrument. It claims that a shared, large-scale, programmable research facility, built from current virtualization and radio-access technologies and managed under the ESFRI lifecycle, would make experimental results in networking reproducible, credible, and reusable. The authors draw on lessons from earlier testbeds and from the open-source 5G ecosystem to sketch the architecture and the full research-lifecycle support, including data management, FAIR principles, and integration with European open-science services, that such an instrument requires. A sympathetic reader would take the paper as a call to treat experimental networking as a science carried out on calibrated shared equipment, with the same seriousness as other experimental disciplines.","feed_headline":"Networking needs a shared instrument for reproducible experiments","feed_subtitle":"A Europe-wide facility would make experimental network results repeatable and credible, the authors argue.","key_machinery":"The central object is SLICES itself, a distributed research infrastructure organized as a central hub plus national nodes, each node built from four subsystems: inter-facility and intra-facility switching, real-time and non-real-time computing, radio infrastructure, and end-user devices. It is a layered architecture comprising Resource, Virtualization, Orchestration, Northbound-Interface, Application, and UI layers, in which a central SLICES-Core application acts as a multi-domain orchestrator that glues NFV, MEC, and cloud-native orchestrators under one authority. The other load-bearing mechanism is the full research-lifecycle methodology, including FAIR data management, metadata, data governance, and reproducibility, which the authors say an instrument must provide rather than just raw testbed capacity.","core_discovery":"The paper's central claim is that experimentally-driven networking research can and should be built on a scientific instrument: a distributed, programmable, pan-European facility called SLICES that supports the whole research lifecycle from experiment design through data collection to publication. The authors argue that previous generations of testbeds, from PlanetLab and ORBIT to GENI and FIRE, demonstrated both the value and the limits of federated platforms, and that the current availability of SDN/NFV, network slicing, disaggregation, and open-source 5G software makes a next-generation instrument feasible. They propose that SLICES should be governed under ESFRI, adopt FAIR data principles, and interoperate with EOSC, so that networking results become reproducible and credible in the same way as results in established sciences.","pith_inferences":["The same reproducibility crisis exists in other computer-science fields; if SLICES works, its lifecycle and governance model could plausibly be copied by distributed-systems or AI/ML communities that also depend on experimental evidence.","The authors' emphasis on the research lifecycle suggests that the value of a testbed may shift from raw hardware to data stewardship; funders could start evaluating facilities by reproducibility metrics rather than capacity.","Whether the facility stays relevant depends on the open-source 5G ecosystem remaining consolidated; a divergence among projects like OAI, srsRAN, and O-RAN implementations would break the architecture's assumption that commodity software can stand in for proprietary network equipment."],"forward_implications":["If SLICES is right, networking journals and venues could come to expect artifacts that run on a shared facility rather than on unverifiable private testbeds.","Beyond-5G and 6G innovations could be tested end-to-end on a reference infrastructure before deployment, the way other fields test on calibrated instruments.","Small research groups and SMEs would gain access to large-scale 5G and edge facilities they cannot build themselves.","Data from experiments would be preserved under FAIR principles, making it possible to re-analyze and combine past experiments rather than repeat them from scratch."],"supporting_citations":[{"why":"Provides the first large-scale federated testbed (PlanetLab Europe/OneLab) whose lessons the paper builds on.","marker":"[1]"},{"why":"Shows a wireless testbed (ORBIT) that succeeded by offering a realistic environment for wireless protocol experiments.","marker":"[4]"},{"why":"Supplies the GENI example of a structured national network-research infrastructure.","marker":"[5]"},{"why":"Supplies the FIRE initiative example of federating heterogeneous European testbeds.","marker":"[6]"},{"why":"Defines the Slice-based Federation Architecture that the paper says could inspire SLICES' multi-domain control.","marker":"[7]"},{"why":"Provides OpenAirInterface, an open-source 5G stack the paper identifies as a core building block for software-defined RAN experiments.","marker":"[14]"},{"why":"Defines the ESFRI framework and lifecycle that SLICES must follow to be a recognized scientific instrument.","marker":"[15]"},{"why":"States the FAIR guiding principles that the paper argues must govern data produced by the instrument.","marker":"[84]"}],"fun_headline_variants":["Networking research gets a shared scientific instrument","SLICES: a testbed for all of Europe's network science","Making network results reproducible with a shared facility","The networking community's move to credible experiments","A scientific instrument for repeatable network research"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument collapses if a large, centrally governed facility planned on a decade-long cycle cannot stay relevant in a field where technology changes quickly and researchers remain fragmented.","fun_headline_variants_meta":{"raw":{"variants":["Networking research gets a shared scientific instrument","SLICES: a testbed for all of Europe's network science","Making network results reproducible with a shared facility","The networking community's move to credible experiments","A scientific instrument for repeatable network research"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00033,"raw_usage":{"total_tokens":1812,"prompt_tokens":891,"completion_tokens":921,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":507,"completion_tokens_details":{"reasoning_tokens":850}},"tokens_in":507,"tokens_out":921,"duration_ms":8639,"temperature":1.0,"reasoning_tokens":850,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T20:30:23.846262+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"See whether the fraction of networking papers whose experiments can be independently reproduced on SLICES rises above today's baseline within five years of the facility opening; if it does not, the paper's central claim is not borne out.","supporting_citations":[{"cited_title":"OpenAirInterface: A flexible platform for 5G research,","cited_arxiv_id":null,"evidence_quote":"Provides OpenAirInterface, an open-source 5G stack the paper identifies as a core building block for software-defined RAN experiments."},{"cited_title":"European Strategy Forum on Research Infrastructures (ESFRI),","cited_arxiv_id":null,"evidence_quote":"Defines the ESFRI framework and lifecycle that SLICES must follow to be a recognized scientific instrument."},{"cited_title":"The FAIR Guiding Principles for scientific data management and stewardship,","cited_arxiv_id":null,"evidence_quote":"States the FAIR guiding principles that the paper argues must govern data produced by the instrument."}],"review_version":1}