Network Slice Embedding over Space Division Multiplexed Elastic Optical Networks
Divya Khanure ∗ * , Riti Gour † \dagger , Congzhou Li ∗ * , and Jason P. Jue ∗ * · 2026
Read Network Slice Embedding over Space Division Multiplexed Elastic Optical Networks by Divya Khanure ∗ * , Riti Gour † \dagger , Congzhou Li ∗ * , and Jason P. Jue ∗ * (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-11936.txt.
Introduction
Khanure, Gour, Li, and Jue tackle network slice embedding over space-division multiplexed elastic optical networks. There, slices need coordinated spectrum and compute resources under shifting traffic. Earlier studies of routing, modulation, core, spectrum, and compute usually treat those dimensions apart. Their Waypoint Assisted Multi Segment Slice Mapping scheme couples compute placement with spectrum allocation in sequence. It improves acceptance ratios by up to 27 percent under high load and cuts total provisioning cost by up to 47 percent against the baseline.
Catalog record
| CREATOR | Divya Khanure ∗ * , Riti Gour † \dagger , Congzhou Li ∗ * , and Jason P. Jue ∗ * |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 37721 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.11936
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Network Slice Embedding over Space Division Multiplexed Elastic Optical Networks Divya Khanure ∗ * , Riti Gour † \dagger , Congzhou Li ∗ * , and Jason P. Jue ∗ * Abstract Network slicing over space-division-multiplexed elastic optical networks (SDM-EONs) enables efficient multiservice provisioning on a shared optical substrate. However, embedding such slices requires coordinated spectrum and compute resource management under dynamic traffic, which most existing RMCSA studies treat independently. This paper focuses on the network slice embedding problem over space-division-multiplexed elastic optical networks (SDM-EONs), aiming to develop efficient resource allocation strategies that ensure both high utilization and reliable service performance. While prior studies have investigated routing, modulation format, core, and spectrum allocation (RMCSA), they typically consider these dimensions separately from compute placement. To address this gap, this paper proposes a Waypoint-Assisted Multi-Segment Slice Mapping (WMSM) scheme, which integrates compute placement with spectrum allocation in a sequential but coupled manner to enable more flexible resource placement and improved spectrum efficiency. Numerical results show that WMSM improves acceptance ratios by up to 27% under high-load conditions, while achieving up to 47% lower total provisioning cost relative to the baseline strategy. These results highlight the benefits of integrated compute–spectrum provisioning and provide design insights for scalable, compute-aware optical slice mapping. Email:{divya.khanure, jjue, congzhou.li}@utdallas.edu, [email protected] I Introduction The exponential growth of bandwidth-intensive and cloud-driven applications continues to push the limits of optical transport infrastructures. Space-division multiplexing (SDM) combined with elastic optical networks (EONs) has emerged as a promising solution, enabling multiple spatial cores per fiber and dynamic spectrum allocation to deliver substantial improvements in capacity, flexibility, and spectral efficiency [ 2 ] . Network slicing further enhances this flexibility by allowing multiple virtual networks to coexist on shared physical resources. In SDM-EONs, slicing spans three dimensions: spectrum, spatial cores, and computational capacity. This introduces coupled constraints — spectrum assignment must satisfy continuity and contiguity, and cores must be selected to mitigate inter-core crosstalk, and compute placement must respect node capacity limits — making joint optimization computationally intractable, while independent treatment leads to poor utilization and increased blocking. Prior studies on network slicing over EONs typically treat compute placement separately from routing, modulation, core, and spectrum assignment (RMCSA), often leading to suboptimal or infeasible slice mappings: spectrum may be reserved along a path where compute capacity is unavailable, or vice versa. Most optical slice-embedding works also assume a strict one-to-one mapping between virtual and substrate nodes, which limits utilization. In the broader VNE literature, Wang et al. [ 14 ] and Chowdhury et al. [ 4 ] showed that node co-location and coordinated node-link mapping improve embedding efficiency and acceptance ratio, yet such strategies have not been extended to optical or SDM-EON slicing frameworks. Our work introduces controlled co-location of slice nodes on shared substrate nodes, improving compute utilization and reducing provisioning cost. To address these gaps, this paper proposes two integrated heuristics for slice mapping in SDM-EONs. Direct Path-Driven Slice Mapping (DPSM) serves as a baseline, prioritizing direct end-to-end routing with sequential compute placement. Waypoint-Assisted Multi-Segment Slice Mapping (WMSM) introduces intermediate compute-capable waypoints and performs segment-wise spectrum allocation, enabling flexible, resource-efficient multi-segment provisioning. Both algorithms couple compute placement and spectrum allocation in a sequential but interdependent manner. The main contributions are: • A coordinated system model for SDM-EON slice mapping that integrates spectrum, compute, and inter-core crosstalk constraints in a sequential but interdependent manner. • A WMSM algorithm evaluated against a DPSM baseline, both with and without balanced placement variants. • A performance evaluation demonstrating significant reductions in blocking probability, spectrum usage, and provisioning cost over baseline prioritization schemes. The rest of the paper is organized as follows: Section II reviews related work. Section III presents the system model and problem formulation. Section IV describes the proposed algorithms. Section V evaluates performance, and Section VI concludes. Fig. 1: E2E network slicing and lightpath provisioning. II Background Literature In single-core EONs, several studies formulated the VNE problem in conjunction with routing and spectrum allocation (RSA). Gong and Zhu [ 5 ] introduced one of the earliest formulations for Virtual Optical Network Embedding (VONE) over EONs, proposing both ILP and heuristic strategies based on layered-auxiliary-graph modeling. Wang and Hu [ 15 ] later improved scalability through a path-growing approach for optical VNE in SLICE networks, achieving near-optimal results while reducing variable complexity. However, these works primarily focused on spectrum utilization and connectivity and did not incorporate spatial division multiplexing or compute-aware placement. Similarly, cross-haul and 5G transport studies, such as Li et al. [ 12 ] and Gu et al. [ 6 ] , integrated slicing and resource allocation across RAN and metro layers, often using heuristic-assisted deep reinforcement learning, but these solutions operate at higher architectural layers and are unsuitable for fast, core-level optical provisioning. With the emergence of multi-core fibers, research on SDM-EONs has addressed the Routing–Core–Spectrum Assignment (RCSA) and Routing–Modulation–Core–Spectrum Assignment (RMCSA) problems, frequently incorporating inter-core crosstalk (IC-XT) constraints. Zhang et al. [ 18 ] and Kumar et al. [ 10 ] proposed crosstalk-aware VONE algorithms that mitigate IC-XT in heterogeneous and spectrally-spatially elastic optical networks, respectively. Jin et al. [ 9 ] further introduced a dynamic topology-aware VONE for hybrid services in SDM-EONs, jointly considering VON topology and physical-layer topology to improve spectral efficiency. Tang et al. [ 13 ] later proposed an ILP model minimizing inter-core crosstalk jointly in spatial, frequency, and time domains for scheduled lightpath demands. More recent RMCSA heuristics, such as Heera et al. [ 8 , 7 ] , introduced congestion-aware and group-processing algorithms for dynamic provisioning and fair spectrum allocation under heterogeneous traffic. While these studies effectively address crosstalk, congestion, and fairness, they treat compute resource placement as independent of optical resource allocation. Extensive research on slice mapping and Virtual Network Embedding (VNE) in optical networks has focused on efficiently mapping virtual networks onto underlying elastic and space-division multiplexed optical infrastructures. Most optical VONE studies adopt a one-to-one node mapping for simplicity and to avoid concentrated compute contention and single-point failures. In contrast, the broader VNE community has examined node co-location as a means of improving embedding efficiency. Wang et al. [ 14 ] proposed the Node-Fusion approach, which evaluates candidate substrate nodes using an integrated ranking metric based on CPU availability, node degree, and adjacent link bandwidth, and introduces a Node-Fusion Interconnection Value (NFIV) to co-locate highly interconnected virtual nodes. This reduces the number of virtual links requiring physical provisioning and minimizes overall embedding cost. Chowdhury et al. [ 4 ] further advanced this idea by coordinating node and link mapping, where node placement decisions directly consider link feasibility and bandwidth cost, enabling more balanced utilization and higher acceptance ratios. While both studies demonstrate that coordinated mapping and node co-location can significantly lower cost and improve resource efficiency, they are confined to abstract VNE models that exclude optical-layer factors such as spectrum continuity, modulation reach, and inter-core crosstalk. To the best of our knowledge, such co-location principles have not been incorporated into optical or SDM-EON slice mapping. Our work bridges this gap by introducing controlled node co-location within an optical slicing framework that sequentially, with balanced placement, accounts for compute constraints, spectrum allocation, and physical-layer impairments under dynamic traffic conditions. This extension preserves the