{"id":"de46d1ad-e6a7-420f-a6e7-e46c8eaec7c1","arxiv_id":"2411.16159","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"The paper formulates a MILP and SWP-based heuristics for RFMSA in hybrid ULL-SSMF elastic optical networks, finding OSNR-aware selection best for static and spectrum-usage-based selection best for dynamic traffic.","lead":"This paper proposes routing, fiber, modulation format, and spectrum allocation algorithms for optical networks that mix ultra-low-loss fiber and standard single-mode fiber on the same link. It reports that an OSNR-aware fiber selection strategy nearly matches an optimal MILP model for static traffic, and that a spectrum-usage strategy lowers blocking probability for dynamic traffic.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Table I lists 16-QAM FS capacity as 700 Gb/s instead of the physical 100 Gb/s for a 12.5 GHz slot; every FS count in Figs. 4–5, and the near-MILP static claim, depends on this erroneous parameter. Verify the value actually used in the simulator.","rationale":"The reader's weakest assumption focused on post-hoc tuning of alpha and Omega. That is a valid generalization concern, but the single most load-bearing issue I see is the concrete parameter error in Table I: the published 16-QAM FS capacity of 700 Gb/s is physically impossible for a 12.5 GHz slot and is inconsistent with the table's own monotone progression. Eq. (1) uses this capacity to compute every demand's FS requirement, so the static maximum-FS comparisons in Figs. 4–5 and the abstract's near-MILP claim all inherit this number. Because both MILP and the SWP heuristics use the same formula, the relative improvement of OA over UFF/Random might survive a correction, but the absolute results and the modulation choices could change substantially. The paper does not disclose the exact value used in the Java implementation, so the reader cannot tell whether the published numbers were produced with 700 or 100. That makes the central quantitative claim unverifiable as written. A simple re-run with 100 Gb/s settles it. I therefore keep the reader's CONDITIONAL verdict: the paper should be revised to correct Table I and to re-validate the static and dynamic results, but there is no evidence yet that the algorithmic framework itself is broken.","tokens_in":17153,"tokens_out":15184,"duration_ms":143497,"concrete_test":"Locate the simulator's FS-capacity table, or re-run the static n6s9 and USNET experiments of Figs. 4–5 with 16-QAM capacity set to 100 Gb/s in Eq. (1), keeping all other settings and the tuned alpha. If the maximum-FS curves shift by more than a few slots, or if the ordering between OA and UFF/Random changes, the published claim is not robust to this parameter. Also re-check the dynamic blocking results, since 16-QAM is used there as well.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section VI.A, Table I gives the FS capacity of 16-QAM as 700 Gb/s. In a 12.5 GHz FS, DP-16QAM carries 100 Gb/s, and the table's own progression (25/50/75/125/150 for BPSK/QPSK/8-QAM/32-QAM/64-QAM) shows 100 is intended. If the published value is used in Eq. (1), any demand up to 700 Gb/s needs only 1 FS under 16-QAM instead of 2–7 FS, artificially collapsing spectrum-consumption differences. Figures 4 and 5, and the claim that OA achieves near-MILP maximum FS usage, are computed with this number. A wrong FS capacity changes which modulation the algorithm prefers and can change the gap between OA and UFF/Random, so the central static result is not reproducible as stated. If the implementation used the corrected 100 Gb/s, then the printed Table I is still a reproducibility defect; if it used 700, all quantitative conclusions are invalid.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the routing, fiber, modulation format, and spectrum allocation (RFMSA) problem in elastic optical networks where each link contains both a standard single-mode fiber (SSMF) and an ultra-low-loss (ULL) fiber. It formulates a node-arc mixed-integer linear programming (MILP) model for static traffic and proposes Spectrum Window Plane (SWP)-based heuristic algorithms with four fiber selection strategies: Random, ULL-fiber-first (UFF), OSNR-aware (OA), and spectrum-usage (SU). Static simulations on a 6-node network (n6s9) and a 24-node USNET show that OA achieves the lowest maximum number of frequency slots (FSs) used, close to the MILP on n6s9; dynamic simulations on both networks show that SU reduces lightpath blocking probability compared with Random and UFF. A cost analysis argues that a hybrid ULL+SSMF deployment is more cost-effective than all-ULL or all-SSMF alternatives.","tokens_in":17371,"tokens_out":7447,"duration_ms":60925,"significance":"If the claims hold, the OA heuristic offers a low-complexity approach to static RFMSA that is near-optimal on small networks, and the SU strategy provides an effective online fiber-selection rule for dynamic traffic. The MILP formulation is a useful benchmark for future work on hybrid-fiber elastic optical networks. The paper also contributes a cost comparison that supports the practical motivation for deploying ULL fibers alongside existing SSMFs. However, the evidence is weakened by parameter tuning on the test networks and by a concrete error in the modulation-format capacity table that percolates into the reported FS counts.","major_comments":[{"comment":"The FS capacity of 16-QAM is listed as 700 Gb/s, which is inconsistent with the 12.5 GHz slot granularity and with the table's own progression (BPSK 25, QPSK 50, 8-QAM 75, 32-QAM 125, 64-QAM 150 Gb/s); the expected value is 100 Gb/s. The example in Fig. 1(b) is also internally inconsistent: a 160 Gb/s demand using 16-QAM is said to require 2 FS, which only holds if E=100 Gb/s, whereas a 180 Gb/s demand using 8-QAM (E=75) is reported as 6 FS, although the formula would give 3. If the simulator actually uses 700 Gb/s, then every demand up to 700 Gb/s occupies exactly 1 FS under 16-QAM, artificially collapsing modulation-dependent differences and invalidating the comparisons in Figs. 4 and 5 and the near-MILP claim. If the corrected value 100 Gb/s was used, the printed table is still a reproducibility defect that must be fixed. Please correct Table I, reconcile the example in Fig. 1, and state explicitly which value was used in the simulations.","section":"Section VI.A, Table I and Eq. (1)"},{"comment":"The OA strategy's threshold alpha is selected post-hoc on the same test networks: Fig. 7(a) shows an optimum at alpha=1.12 for n6s9 and Fig. 7(b) shows an optimum at alpha=1.09 for USNET, and these values are then used to produce the near-MILP results in Figs. 4 and 5. Because alpha is tuned on the evaluation networks, the reported advantage of OA over UFF and Random may be optimistic and may not generalize to other topologies or traffic patterns. The authors should either fix alpha by a general rule (e.g., derived from fiber loss parameters), evaluate on a third network, or report the performance range over a realistic alpha interval.","section":"Section VI.C, alpha threshold"},{"comment":"The SU strategy uses a cost function with a hand-chosen weight Omega (1.2 for ULL schemes that reduce FS count, 0.8 otherwise, 1 for all-SSMF schemes) and no sensitivity analysis is provided. The paper's central dynamic-traffic claim—that SU remarkably surpasses UFF and Random in blocking probability—rests entirely on this fixed weight. Please show that the blocking-probability advantage is robust to Omega over a plausible range (e.g., 1.0 to 1.5) or justify the chosen value from first principles.","section":"Section V.A, Eq. (34) and Section VI.B.2"}],"minor_comments":[{"comment":"Equation (18) appears to contain a typo: it should use E_{sd1,ij} and E_{sd2,ij} for the two distinct lightpaths, not the same variable twice.","section":"Section IV, Eq. (18)"},{"comment":"The notation B_h^m in Eq. (34) is confusing; from the numerical example, the formula is c_h = n_h^d * B_h / (beta - 1) * Omega, where beta-1 is the maximum number of adjacent FS state changes. Please rewrite the equation to match the example.","section":"Section V.A, Eq. (34) and Fig. 2"},{"comment":"The phrase 'exhibits optimal performance' in the abstract is overstated for a heuristic; the simulation only shows that OA is close to the MILP on a small network.","section":"Abstract and Section VI.B.1"},{"comment":"The total fiber cost values listed in the text (8584, 85840, 171680 units) do not clearly follow from the stated formula and the four deployment scenarios; please reconcile the numbers and specify which scenario corresponds to each value.","section":"Section VI.D, cost analysis"},{"comment":"The paper refers to Ref. [24] for the OSNR calculation but does not specify key simulation parameters such as amplifier noise figure, span loss, or launch power; please state these values so the results are reproducible.","section":"Section VI.A, OSNR model"},{"comment":"The near-MILP claim should be scoped explicitly to the n6s9 network with the tested demand range, since no MILP results are given for USNET.","section":"Abstract and conclusion"}],"recommendation":"major_revision","confidential_remarks":"The paper is a reasonable extension of the authors' prior work, but the Table I error and the post-hoc tuning of alpha are the two main obstacles. The Table I error is literally load-bearing: if the published value is used, all FS-count results are invalid; if the corrected value was used, the manuscript is still not reproducible as written. The alpha tuning means the headline static-traffic result is not convincingly general. I would like to see a corrected table, a re-run or explicit confirmation of the simulator value, a sensitivity analysis for alpha and Omega, and ideally a validation on a third network before publication."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is a credible, incremental extension of SWP-based RMSA to hybrid ULL/SSMF links, with a MILP benchmark and two new fiber-selection ideas. It deserves serious refereeing, but the static results as printed are not reproducible because Table I lists 16-QAM FS capacity as 700 Gb/s instead of 100 Gb/s.\n\nThe genuinely new piece is the RFMSA formulation on links where ULL and SSMF coexist, which is absent from the cited multi-core/multi-fiber literature. The MILP is a solid node-arc model, and the OA and SU strategies are sensible, with the SU cost formula being a reasonable way to balance spectrum fragmentation and fiber quality. The authors also compare against their earlier Adaptive algorithm and give a useful complexity reduction.\n\nThe soft spots are real but not fatal. The OA threshold alpha is tuned on the exact test networks (1.12 for n6s9, 1.09 for USNET), and there is no evaluation on unseen networks or a fixed rule. The SU weight Omega is hand-chosen. The MILP comparison only covers the small n6s9 network with modest demands. More important, Table I has the 700/100 Gb/s error. The Fig. 1 example uses 100 (160 Gb/s -> 2 FS), so either the simulator used 100 and the table is a typo, or the simulator used 700 and the example is wrong. Either way, the printed record is inconsistent, and since Eq. (1) converts demand to FS count, the static curves in Figs. 4–5 depend on this value. The authors must state which value was used and re-run or confirm the results. The example also has a separate arithmetic slip (180 Gb/s at 8-QAM gives 3 FS, not 6), which suggests the manuscript was not carefully proofread. None of this is a conceptual flaw, but it blocks verification.\n\nWho this is for: people working on elastic optical network planning with mixed fiber types; the paper is a reasonable starting point for that scenario. I would cite it after the numbers are fixed. If the corrected simulator still shows OA near-MILP, the contribution stands as a modest practical improvement.\n\nRecommendation: send to peer review, but the referee should require a corrected table, a clear statement of the FS capacity used, a fixed or validated alpha, and ideally a comparison on a second dynamic baseline.","headline":"A workmanlike extension of SWP-based RMSA to hybrid ULL/SSMF links with a new MILP benchmark, but the printed static results hinge on a wrong FS-capacity value in Table I that needs author clarification.","tokens_in":17941,"tokens_out":4465,"would_cite":false,"duration_ms":35891,"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":"Per-link fiber choice gets near-optimal spectrum use in hybrid ULL-SSMF networks.","keywords":["elastic optical networks","hybrid ULL fiber-SSMF","RFMSA","spectrum window plane","fiber selection strategy","OSNR-aware strategy","lightpath blocking probability","mixed integer linear programming"],"falsifier":"Run the SWP-based RFMSA algorithm with the OA strategy on a third network, or on the same topologies with altered link lengths, using a fixed $\\alpha$ such as 1.12 or 1.09, and compare the maximum number of FSs used against the MILP bound or against UFF and Random. If the gap widens to the level seen for UFF or Random rather than remaining MILP-like, the central static-traffic claim fails. Similarly, re-running the dynamic experiments with $\\Omega$ fixed at 1 and checking whether SU still outperforms Random and UFF would test the weight's role.","tokens_in":1883,"feed_emoji":"🌐","tokens_out":2976,"duration_ms":90857,"temperature":0.7,"pith_summary":"This paper studies what happens when network operators add ultra-low-loss (ULL) fiber beside existing standard single-mode fiber (SSMF) on every link, so each lightpath must choose a fiber type per link as well as a route, modulation format, and spectrum block. It formulates that joint routing, fiber, modulation format, and spectrum allocation (RFMSA) problem as a node-arc MILP for static traffic and builds Spectrum Window Plane heuristics around four fiber-selection strategies. The central claim is that an optical signal-to-noise ratio (OSNR)-aware strategy, which picks the ULL fiber only when its OSNR advantage over SSMF on a link exceeds a threshold, achieves the same maximum frequency-slot usage as the MILP model in static scenarios, while a spectrum-usage strategy that scores fiber-selection schemes by available contiguous capacity markedly reduces lightpath blocking under dynamic traffic. The paper further argues that a hybrid one-ULL-plus-one-SSMF deployment delivers nearly the same blocking performance as two ULL fibers per link at half the fiber cost. The contribution is a near-optimal operating rule for the transition period when ULL fiber is added alongside, rather than replacing, existing fiber.","feed_headline":"Per-link fiber choice gets near-optimal spectrum use","feed_subtitle":"In mixed ULL-SSMF networks, an OSNR-aware rule matches MILP spectrum use while a spectrum-usage rule cuts blocking.","key_machinery":"The central mechanism is the Spectrum Window Plane (SWP): for each modulation format, a layered auxiliary graph where a layer is a spectrum window of contiguous free frequency slots, and a virtual link exists on a physical link only if the chosen fiber has that window free; Dijkstra on the SWP finds the shortest route. Around this, the paper wraps a node-arc MILP model, whose objective is to minimize the largest FS index used anywhere in the network, and four fiber-selection strategies. The OSNR-aware strategy maps ULL to the SWP when the ratio of ULL to SSMF OSNR exceeds a threshold $\\alpha$, otherwise SSMF; the SU strategy computes, for each of the $2^n$ fiber-selection schemes along a route, a cost $c_h = n_h^d \\cdot B_h^m \\cdot \\Omega$ that rewards abundance of contiguous idle slots and penalizes fragmentation, then picks the highest-cost scheme. The thresholds $\\alpha$ (1.12 for n6s9, 1.09 for USNET) and the weight $\\Omega$ (1, 1.2, or 0.8) are the parameters that carry the claimed near-optimality.","core_discovery":"The discovery is that per-link fiber selection is a main lever for spectrum efficiency in a hybrid ULL-SSMF network, and that a simple local rule can capture most of the benefit. In static traffic, the SWP-based RFMSA algorithm with the OA strategy achieves a maximum number of frequency slots used that is essentially the same as the MILP model's optimum on the n6s9 network, and on USNET it cuts the maximum FS usage by up to 41.7% relative to the ULL-fiber-first strategy and 26.2% relative to the random strategy. In dynamic traffic, the SU strategy, which evaluates the $2^n$ fiber-selection schemes along a candidate route through a cost that rewards more available spectrum blocks, higher adjacency-change counts, and ULL usage that lowers FS demand, yields markedly lower lightpath blocking probability than Random and UFF with both SWP-based and shortest-path routing. The paper further claims that the HUS-EON reaches blocking probabilities close to a two-ULL-fiber network at roughly half the fiber deployment cost.","pith_inferences":["Beyond the paper's claims: if the link threshold $\\alpha$ is set by a general calibration rule rather than tuned per network after seeing results, the OA strategy might become a deployable automatic control; the paper does not establish such a rule.","The SU cost formula is effectively a fragmentation score with a ULL bonus; a parameter-free variant that sets $\\Omega$ from the actual reduction in FS demand caused by choosing ULL on that route could be benchmarked against the reported blocking results.","A broader design principle suggested by the results: in mixed-media networks, fiber selection and spectrum assignment should be optimized jointly, because choosing the low-loss fiber changes both the modulation format, and therefore the number of FSs, and the fragmentation pattern on the link.","The cost comparison assumes normalized fiber prices of 1 and 10 units/km; if ULL fiber prices fall relative to SSMF, the hybrid deployment's cost advantage narrows, and a sensitivity analysis over the price ratio would map the crossover to a dual-ULL network."],"forward_implications":["Static network operators can use the OA strategy with a per-network threshold to provision all demands with near-optimal spectrum usage, measured by the maximum FS index, without solving the MILP.","Dynamic operators can use the SU strategy to reduce lightpath blocking probability substantially compared with always preferring ULL or choosing fibers randomly, under both SWP-based and shortest-path-based routing.","Deploying one ULL fiber alongside each SSMF gives most of the blocking-probability benefit of two ULL fibers per link, at about half the fiber cost.","The OA strategy has lower computational complexity than the earlier adaptive approach, $O(|M|\\,|F|\\,|N|^2\\,|L|)$ versus $O(|M|\\,|F|^2\\,|N|^2\\,|L|\\,|T|)$, making online scheduling in larger networks more feasible.","Allowing a lightpath to use different fiber types on different links can reduce maximum FS usage compared with using a single fiber type end-to-end, as the paper's four-node example illustrates."],"supporting_citations":[{"why":"Prior adaptive SWP-based algorithm for the hybrid ULL fiber-SSMF network that the OA strategy matches in performance and improves on in complexity.","marker":"[8]"},{"why":"Supplies the OSNR calculation formulas used to evaluate candidate routes and select modulation formats.","marker":"[24]"},{"why":"Original SWP-based node-arc heuristic approach that the proposed RFMSA algorithm extends.","marker":"[25]"},{"why":"Datasheet for the ULL fiber used in simulations, providing the 0.166 dB/km attenuation coefficient.","marker":"[26]"},{"why":"Source of the frequency-slot capacities and OSNR thresholds for the six modulation formats in Table I.","marker":"[27]"}],"fun_headline_variants":["Hybrid ULL-SSMF: local rules match MILP spectrum use","Static and dynamic fiber pick: OSNR-aware tops static, SU cuts blocking","Per-link fiber choice halves cost, near-full performance","OSNR-aware fiber allocation hits optimum in static, SU wins dynamic"],"cache_read_input_tokens":20096,"weakest_assumption_plain":"The near-MILP performance of the OSNR-aware strategy rests on the threshold $\\alpha$ being tuned separately for each test network after seeing its results; if a single fixed $\\alpha$ or a rule derived without post-hoc tuning is required, the claimed advantage over simpler strategies is not established by the paper.","fun_headline_variants_meta":{"raw":{"variants":["Hybrid ULL-SSMF: local rules match MILP spectrum use","Static and dynamic fiber pick: OSNR-aware tops static, SU cuts blocking","Per-link fiber choice halves cost, near-full performance","OSNR-aware fiber allocation hits optimum in static, SU wins dynamic"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000305,"raw_usage":{"total_tokens":1812,"prompt_tokens":1067,"completion_tokens":745,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":683,"completion_tokens_details":{"reasoning_tokens":668}},"tokens_in":683,"tokens_out":745,"duration_ms":7997,"temperature":1.0,"reasoning_tokens":668,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T13:29:06.135303+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the SWP-based RFMSA algorithm with the OA strategy on a third network, or on the same topologies with altered link lengths, using a fixed $\\alpha$ such as 1.12 or 1.09, and compare the maximum number of FSs used against the MILP bound or against UFF and Random. If the gap widens to the level seen for UFF or Random rather than remaining MILP-like, the central static-traffic claim fails. Similarly, re-running the dynamic experiments with $\\Omega$ fixed at 1 and checking whether SU still outperforms Random and UFF would test the weight's role.","supporting_citations":[{"cited_title":"Lightpath provisioning in a hybrid ULL fiber -SSMF elastic optical network,","cited_arxiv_id":null,"evidence_quote":"Prior adaptive SWP-based algorithm for the hybrid ULL fiber-SSMF network that the OA strategy matches in performance and improves on in complexity."},{"cited_title":"Ultra-low loss fiber deployment in elastic optical networks with fixed and variable topologies,","cited_arxiv_id":null,"evidence_quote":"Supplies the OSNR calculation formulas used to evaluate candidate routes and select modulation formats."},{"cited_title":"Novel node -arc model and multi -iteration heuristics for static routing and spectrum assignment in elastic optical networks,","cited_arxiv_id":null,"evidence_quote":"Original SWP-based node-arc heuristic approach that the proposed RFMSA algorithm extends."},{"cited_title":"Corning TXF optical fiber,","cited_arxiv_id":null,"evidence_quote":"Datasheet for the ULL fiber used in simulations, providing the 0.166 dB/km attenuation coefficient."},{"cited_title":"Effect of channel launch power on fill margin in C+L band elastic optical networks,","cited_arxiv_id":null,"evidence_quote":"Source of the frequency-slot capacities and OSNR thresholds for the six modulation formats in Table I."}],"review_version":1}