Source-linked AI summary
Potential impact of CV-QKD integration on classical WDM network capacity
Cédric Ware, Raphaël Aymeric, Chaima Zidi, Mounia Lourdiane
TL;DR
Raman noise constrains CV-QKD and classical WDM coexistence, with power limits potentially reducing classical network capacity. The paper uses network-planning simulations to examine this trade-off. Simulations show classical blocking can emerge when CV-QKD traffic exceeds a threshold, motivating dedicated planning heuristics and tools.
Problem
The paper examines how Raman noise and optical-power limits affect the coexistence of CV-QKD and classical WDM traffic in a network.
Source: §I · p. 1Method
Network-planning simulations allocate CV-QKD traffic over a scaled mesh network before classical traffic, using an opaque routing heuristic.
Source: B. Network simulation, topology and… · p. 2, C. CV-QKD traffic routing heuristic:… · p. 2Results
With CV-QKD traffic, classical blocking develops a floor above a certain offered-QKD level, while the no-QKD classical network capacity is about 20 Tbit/s.
Source: B. Classical capacity vs CV-QKD · p. 3Takeaways & Limitations
Mixed classical and CV-QKD networks will require dedicated planning heuristics and tools to find better compromises between the traffic types.
Source: §IV · p. 4, §V · p. 4The results are preliminary and use one small topology; the Raman-noise model does not accurately represent physical Raman-noise generation.
Source: §IV · p. 4
Abstract
from arXiv · showhide
Continuous-variable quantum key distribution (CV-QKD) could allow QKD and classical optical signals physically sharing the same optical fibers in existing networks. However, Raman scattering imposes a limit on the optical power, which in turn impacts the network capacity for classical traffic in presence of CV-QKD. Network-planning simulations indicate that maxing out the CV-QKD capacity in an optical link can adversely impact its classical capacity. Although preliminary, these results show that designing a mixed classical and CV-QKD network will require dedicated planning heuristics and tools that specifically seek a compromise between classical and CV-QKD traffics.
I. INTRODUCTION
CV-QKD can coexist with classical WDM traffic, but Raman scattering constrains that coexistence and may reduce classical network capacity. This paper studies the issue at network scale, beyond prior single-link coexistence analyses.
I. INTRODUCTION
- Prior studies examined CV-QKD key rates and coexistence with intense WDM channels on single links, while quantum-network design work did not address coexistence.The paper therefore searches for coexistence limits within a network.Source: §I · p. 1
- Raman scattering creates crosstalk-like noise between WDM channels, while limiting optical power to protect CV-QKD also caps the classical WDM channels a fiber can carry.The paper's simulations find that above a threshold of offered CV-QKD traffic, some optical links cannot sustain classical traffic, increasing its blocking probability.Source: §I · p. 1
- The study uses network-planning simulations to examine coexistence limits and discusses preliminary results, their limits, and further investigation.The paper concludes that mixed classical and CV-QKD networks require dedicated planning heuristics and tools.Source: §I · p. 1
II. METHODOLOGY AND PARAMETERS
The model combines classical WDM network simulation with CV-QKD key-rate calculations, using assumed physical and technological parameters. It also specifies a uniformly scaled network topology and optimistic CV-QKD receiver and noise settings.
II. METHODOLOGY AND PARAMETERS
- The simulations use assumed classical-network and CV-QKD physical and technological parameters, with the CV-QKD key rate calculated in the asymptotic regime.The rate is K = fsym · r, where fsym is symbol rate and r is secret fraction; its optimal value depends on transmittivity T, excess noise ξ, reconciliation efficiency β, and detection type µ.Source: A. CV-QKD and classical signal… · p. 1
- The model assumes a 100-% efficient receiver and link loss combining fixed insertion loss with fiber attenuation.The topology distances are uniformly scaled by factor Λ.Source: A. CV-QKD and classical signal… · p. 2, Figure 1 · p. 2
- The model uses ξ0 = 10−3 SNU for untrusted detection noise and fsym = 1 Gbaud, both described as optimistic choices.The authors identify the choice of Raman efficiency ξR as a major uncertainty for further investigation.Source: A. CV-QKD and classical signal… · p. 2
B. Network simulation, topology and traffic matrix
The study simulates a seven-node Spanish mesh network with 42 demands, scaling link distances and offered traffic to examine network planning under different loads. CV-QKD demands are allocated before classical demands.
B. Network simulation, topology and traffic matrix
- The simulations use Net2Plan in offline design mode on a fictitious mesh connecting seven Spanish cities, with 16 unidirectional links and 42 demands.The 42 demands offer between 6.95 and 1815 Gbit/s, totaling 10 Tbit/s; links average 273 km before distance scaling.Source: B. Network simulation, topology and… · p. 2
- Uniform distance scaling Λ ranges from 0.01 to 0.25 to represent shorter networks, while the same traffic matrix is used for classical and CV-QKD demands.At Λ = 0.01, the longest link is 3.9 km and supports 345 Mbit/s of CV-QKD; at Λ = 0.05 it is 19.6 km and supports 88 Mbit/s, and at Λ = 0.1 it is 39.1 km and supports none.Source: B. Network simulation, topology and… · p. 2
- CV-QKD is allocated first and receives priority over classical traffic so the simulations can study the impact of a given QKD load on classical capacity.Offered traffic is uniformly scaled to vary network load.Source: B. Network simulation, topology and… · p. 2
C. CV-QKD traffic routing heuristic: opaque
The opaque routing heuristic avoids optical bypass and assumes trusted relays, allocating CV-QKD demands over paths with available link capacity. An optional margin below maximum capacity can affect routing outcomes.
C. CV-QKD traffic routing heuristic: opaque
- CV-QKD uses opaque routing without optical bypass, with each node acting as a trusted relay that aggregates demands sharing a link.Bypass is omitted because it would lengthen paths and reduce capacity.Source: C. CV-QKD traffic routing heuristic:… · p. 2
- For each link, the heuristic calculates maximum key-rate capacity without Raman noise, then serves the largest unallocated demand using the best-capacity option among k shortest paths.Each route carries as much of that demand as possible without exceeding any link's remaining capacity; the process repeats until demands are served or no more routes can be added.Source: C. CV-QKD traffic routing heuristic:… · p. 2
- An optional 1 % margin keeps CV-QKD traffic below each link's maximum capacity and is reported to have a strong impact.The margin is applied to the aggregated CV-QKD traffic on each link.Source: C. CV-QKD traffic routing heuristic:… · p. 2
D. Classical traffic RWA heuristic: transparent fixed-grid
The classical traffic heuristic routes demands as transparent-network lightpaths, preserving wavelength continuity along each path. It adds lightpaths only when a WDM slot is available end-to-end and the route preserves each link’s CV-QKD capacity.
D. Classical traffic RWA heuristic: transparent fixed-grid
- The RWA heuristic creates direct lightpaths that occupy the same WDM channel across every link in a path.It assumes each node is a flexible ROADM; each lightpath occupies one WDM slot and carries fixed traffic R.Source: D. Classical traffic RWA heuristic:… · p. 2
- The heuristic does not require multiple lightpaths for a large demand to share a route or occupy contiguous WDM slots.Regeneration is not considered, as all paths are treated as classically feasible given the short ranges.Source: D. Classical traffic RWA heuristic:… · p. 2
- Demands are served in decreasing order of unallocated traffic, testing the k shortest paths for an available end-to-end WDM slot.A 100 Gbit/s lightpath is added over the shortest path that also preserves every link’s CV-QKD capacity; the process stops when demands are served or no lightpaths can be added.Source: D. Classical traffic RWA heuristic:… · p. 2
III. SIMULATION RESULTS
Simulations show that CV-QKD capacity with negligible blocking is about 0.7–2 Gbit/s at Λ = 0.01–0.05, falling to 100–200 Mbit/s at Λ = 0.1. Figure 2 reports blocking, link use, and utilization across network load, distance scaling, Raman efficiency, and CV-QKD capacity margins.
A. CV-QKD capacity vs network scale
- With negligible blocking, network capacity is about 0.7–2 Gbit/s for Λ = 0.01–0.05, then drops to 100–200 Mbit/s for Λ = 0.1; above Λ = 0.1, distances are too large for significant traffic.Source: A. CV-QKD capacity vs network scale · p. 2
- Panels (a)–(e) plot blocked traffic against offered traffic across distance scaling factors, Raman efficiencies, and, in some plots, a 1-% CV-QKD capacity margin.The caption defines low Raman as ξR = 10−12 SNU/W and high Raman as 10 SNU/W.Source: Figure 2 · p. 3
- Panels (f)–(i) show unused classical links and average link utilization against offered CV-QKD traffic or Raman efficiency.Panels (h)–(i) compare utilization without a capacity margin and with margins from 0 to 10-%.Source: Figure 2 · p. 3
B. Classical capacity vs CV-QKD
With no CV-QKD capacity margin, CV-QKD traffic can produce classical blocking even at low classical loads, with blocking rising in steps as CV-QKD traffic increases. The results associate these steps with links becoming unavailable to classical traffic when they are loaded to maximum CV-QKD capacity.
B. Classical capacity vs CV-QKD
- About 20 Tbit/s is the classical network capacity without CV-QKD, but high-Raman operation reduces it to about 3 Tbit/s even with 1 kbit/s of CV-QKD traffic.In the low-Raman case, classical blocking is unchanged below a threshold of CV-QKD traffic, then a blocking floor appears even at very low classical traffic.Source: B. Classical capacity vs CV-QKD · p. 3
- In Fig. 2(e), classical-traffic blocking increases in steps as offered CV-QKD traffic rises; even the spurious spike at 1.6 · 10^8 bit/s for Λ = 0.1 reaches the later plateau’s level.Source: B. Classical capacity vs CV-QKD · p. 3
- As CV-QKD traffic increases, links stop carrying classical traffic; utilization can first rise as classical traffic takes more circuitous routes, then fall as links become unavailable.With 10 Tbit/s of offered classical traffic, utilization starts at 13%, compared with 16% for offered traffic divided by theoretical maximum capacity.Source: B. Classical capacity vs CV-QKD · p. 3
- The range where classical impact varies most, 10^8–10^10 bit/s of CV-QKD traffic, coincides with rising CV-QKD blocking.The authors suggest that links carrying their maximum CV-QKD capacity become unavailable to classical traffic because adding optical power reduces CV-QKD capacity below the required level.Source: B. Classical capacity vs CV-QKD · p. 3
- A 1-% margin eliminates CV-QKD’s impact on classical traffic in the low-Raman case, while having no effect in the high-Raman case.The low-Raman curves overlap the no-QKD curve; the high-Raman graphs remain identical to the no-margin case.Source: B. Classical capacity vs CV-QKD · pp. 3–4
C. Choice of Raman noise efficiency
The simulations use ξR = 10−12 SNU/W and 10 SNU/W to represent a Raman-efficiency plateau and a regime where classical-link impacts worsen. A possible numerical artifact at very low ξR and the dependence of margin effects on ξR qualify those choices.
C. Choice of Raman noise efficiency
- The model treats Raman noise efficiency ξR as a simplified assessment of Raman nonlinear scattering.The authors say they will test empirically whether a more realistic model is needed at this stage.Source: C. Choice of Raman noise efficiency · p. 4
- At ξR ≤ 10−16 SNU/W, average link utilization stays high, but the authors warn this may be a floating-point artifact.They ignore that regime as Raman noise is considered unlikely to be so low.Source: C. Choice of Raman noise efficiency · p. 4
- The authors select ξR = 10−12 SNU/W on the 10−15–10−1 plateau and 10 SNU/W in the higher-efficiency regime where impacts worsen.Simulations at ξR = 10−3 SNU/W produced the same results as at 10−12 SNU/W.Source: C. Choice of Raman noise efficiency · p. 4
- A CV-QKD capacity margin can keep link utilization high up to an ξR threshold that depends on the margin and possibly CV-QKD traffic.The authors suggest that a margin helps maintain classical capacity in some cases at low Raman efficiency, but not, for example, with the lowest CV-QKD traffic.Source: C. Choice of Raman noise efficiency · p. 4
IV. DISCUSSION AND QUESTIONS FOR FUTURE WORK
The preliminary results indicate that CV-QKD can significantly affect classical capacity in some cases, particularly when Raman noise is high. The discussion identifies unresolved limits in topology, capacity margins, and Raman modeling, and points to planning heuristics that seek a compromise between classical and CV-QKD traffic.
IV. DISCUSSION AND QUESTIONS FOR FUTURE WORK
- The study is preliminary, and the capacity impact is case-dependent, especially when Raman noise is high.The authors describe this as opening several lines of investigation.Source: §IV · p. 4
- Only one small topology was studied, leaving open whether more densely meshed networks’ alternate paths compensate for capacity loss.The authors also ask which links become unavailable first, suggesting longer links may be more vulnerable.Source: §IV · p. 4
- Limiting CV-QKD traffic slightly below each link’s maximum capacity seems to help, but the required margin remains unknown.The margin depends on Raman efficiency, which may require a more realistic model.Source: §IV · p. 4
- Advanced planning heuristics and design tools could seek better compromises between classical and CV-QKD traffic, including priority management between their traffic matrices.Source: §IV · p. 4
- The linear Raman-noise model may show that an impact exists, but it does not accurately represent physical Raman-noise generation.The needed accuracy is unresolved; the required margin depends on Raman noise and spectral distance may determine whether nearby WDM wavelengths or a separate band are needed.Source: §IV · p. 4
V. CONCLUSION
Network-planning simulations show that sharing optical fibers can in some cases adversely affect classical capacity because CV-QKD traffic must be protected from excessive Raman noise. A capacity margin may help, but its required size remains open, and mixed networks appear to need dedicated planning heuristics and tools.
V. CONCLUSION
- In some cases, coexistence adversely affects classical network capacity because CV-QKD traffic must be protected from excessive Raman noise.Source: §V · p. 4
- Limiting CV-QKD traffic slightly below each optical link’s maximum capacity is a possible precaution, but the necessary margin remains an open question.Source: §V · p. 4
- Designing a mixed classical and CV-QKD network appears to require dedicated planning heuristics and tools.Source: §V · p. 4