Traffic Chunk Sizing vs. Optical Switching Speed in Future All-Optical Satellite Networks Sleman Mouammar, Thomas Röthig, Soheil Hosseini, Ítalo Brasileiro, and Admela Jukan
arXiv:2605.04829v1 [cs.NI] 6 May 2026
Institut für Datentechnik und Kommunikationsnetze, Technische Universität Braunschweig, Germany.
Abstract—To enable efficient resource utilization under stringent Size, Weight, and Power (SWaP) constraints through transparent and all-optical switched satellites transmission, various switching paradigms can be considered, including packet, burst, or circuit. To this end, the traffic assembly and algorithmic design for path computations at the ground stations play a key role in determining the switching fabric design. Generally, traffic can be buffered and assembled in chunks at the ground stations and forwarded over the pre-computed optical path in space, similar to terrestrial optical burst switching or fast circuit switching. Regardless of the chosen paradigm, the switching fabric must satisfy specific latency performance requirements. This paper studies the performance of all-optical satellite networks based on the maximum traffic chunk sizes that can be scheduled and the performance of optical switching fabrics in the future over all-optical constellations. We consider various optical switching technologies, including MEMS- and integrated photonic-based solutions, in the context of switching speed, power consumption, and insertion loss. Simulation results indicate that traffic chunk size critically impacts the performance required by optical switching fabrics onboard a satellite. Index Terms—All-optical switching, satellite communication, chunk size, switching fabrics
I. I NTRODUCTION Optical satellite networks are emerging as a key innovation frontier for extending the capacity and coverage of current terrestrial networks. Modern satellite constellations, such as Starlink and Telesat, reportedly employ laser inter-satellite links (LISLs) together with packet-based traffic [1]. As transmission rates scale to hundreds of Gb/s, all-optical switching paradigms are expected to emerge, drawing on approaches established in terrestrial networks such as optical burst switching (OBS), optical packet switching (OPS), and optical circuit switching (OCS) [2]. Regardless of the paradigm employed, the deployed switching fabric must satisfy stringent latency requirements. The fundamental challenge is therefore to enable effective transmission over all-optical paths in modern constellations while preserving low end-to-end (E2E) latency, which is one of the salient advantages of satellite networks. Fig. 1 illustrates this challenge. In ground stations, ISP traffic needs to be assembled in traffic chunks and stored for scheduled satellite transmission. Parameters such as the number of orbital planes and satellites, constellation layout (e.g., Walker delta or Walker star), and orbital altitude have a direct impact on the propagation delay and the E2E latency. For every traffic chunk to be transmitted, routing and wavelength allocation procedures are performed, as well as scheduling in
Fig. 1: Traffic chunk sizing and onboard optical switching in next-generation all-optical satellite networks.
case no immediate transmission is feasible. As illustrated, a wavelength division multiplexing (WDM) input allocates one of the wavelengths, from λ1 to λ4 . Finally, the computed paths need to be reserved for a transmission to start. Onboard the satellite, the optical switching fabric directs the signal towards one of the outputs, with its related physical impairments. In this paper, we study the maximum size of the traffic chunk to be transmitted over the all-optical satellite network, for a given E2E delay constraint. The paper considers three representative LEO constellations: Telesat, Amazon LEO Shell 1, and Starlink Phase 1. It evaluates the impact of different traffic chunks for different onboard optical switching devices, balancing switching speed, power consumption, and network efficiency. We model the switching mechanism similar to OBS, however, with some important differences, including out-of-band control known from OCS, hence the name chunk. We evaluate the related switching parameters, ensuring that switching performance is aligned with traffic chunk sizing and E2E delay. The results show that the maximum feasible traffic chunk for ns-optical is around 500MB for various constellations, whereas for ms-switch, the maximum burst size is significantly lower. The rest of the paper is organized as follows. Section II provides related work. In Section III, we present a reference architecture. Section IV presents the simulation setup and performance evaluation. Finally, Section V concludes the paper.
II. R ELATED W ORK Although optical switching paradigms have been extensively investigated in terrestrial networks, their adoption in satellite systems remains relatively limited. Recent work examines individual important mechanisms. Wang et al. [3] introduce a quality of service (QoS)-adaptive burst assembly algorithm for space-based OBS networks, reducing assembly delay for highpriority traffic by up to 14.68%. Similarly, [4] proposes an optical switched satellite backbone network combining timedivision multiplexing with OBS to enable QoS-aware resource allocation, achieving low latency and near-zero packet loss under specific system configurations. Neither of these works, however, relates the burst formation to hardware constraints and switching-related aspects. In parallel, advances in laser communication and onboard optical switching eliminate OEO conversion, thereby reducing latency, power consumption, and congestion [5]. However, feasible engineering and realization depend critically on the underlying switching fabric. Technologies such as MicroElectro-Mechanical Systems (MEMS) and integrated photonic platforms offer distinct trade-offs in switching speed, power consumption, insertion loss, and system complexity [6]. While modern optical switches can achieve sub-microsecond to nanosecond reconfiguration times with reduced power consumption, prior work focused on device-level performance, without evaluating system-level implications. Experimental efforts have started to address this gap. Zhai et al. [7] validate an OBS-based satellite payload using an arrayed waveguide grating router, demonstrating transparent optical switching with wavelength conversion capabilities. However, their system is limited by a reconfiguration interval of 11.9 µs and relies on wavelength conversion, which may not be compatible with strict SWaP constraints. From a technological perspective, MEMS-based switches are mature and suitable for large-scale deployments due to
high port density and optical transparency. but they typically suffer from relatively slow switching times (ms range) and larger physical dimensions. On the other hand, integrated photonic switches enable ns-scale reconfiguration and compact integration, making them attractive for size- and weightconstrained satellite nodes. Their ultrafast switching capability is particularly advantageous in OBS and fast circuit switching environments, where reduced reconfiguration time directly improves performance. Table I surveys different switching technologies studied in this paper, including commercial MEMS-based (GLSUN) [8], electro-optic (AGILTRON) [9], and beam-steering solutions (POLATIS) [10], as well as an emerging Indium Phosphide (InP) integrated photonic switch, currently assessed as TRL 4 [11]. TABLE I: Optical switching fabrics: performance comparison POLATIS GLSUN AGILTRON InP SOA-based (Beam-steering) (MEMS) (Electro-optic) (PIC) Reference [10] [8] [9] [11] Speed 25 ms 8 ms 0.1 µs 5.2 ns Power Cons. 5W 1.25 W 10 W 0.58 W Insertion Loss 1 dB 2.6 dB 3.5 dB 0 dB TRL 9 9 9 4 Feature
III. R EFERENCE ARCHITECTURE Fig. 2 illustrates the reference architecture, which is detailed in the following subsections: System Setup, introducing the constellation and traffic design; Algorithms, presenting the routing, resource reservation, and scheduling procedures; and Launching the Chunk, describing the transmission of the control packet and data chunk. A. System setup In ground stations, traffic is assembled into chunks and eventually buffered for a scheduled transmission. The first setup component to consider is the constellation configuration
Fig. 2: Reference architecture.
(A). The topological aspects of satellite networks, such as altitude and density, have high impact on the overall E2E latency. In order to cover constellations with different dimensions, this work evaluates three Walker-Delta LEO satellite constellations as illustrated in Table. II: one high-altitude configuration with fewer satellites (Telesat), which offers longer uplinks and downlinks windows, but incurs higher propagation delays, and two lower-altitude, denser constellations, namely Amazon Leo shell 1 and Starlink phase 1, the latter being more dense, to assess the impact of satellite constellation density on system behavior. In all constellations, each satellite node is equipped with four ports for communication with neighboring satellites, arranged in a Grid+ topology, along with an additional port for optical ground station (OGS) connectivity. Based on the Input, the Constellation Three-Line Element (TLE) file (A) contains mean orbital elements derived from tracking data and is formatted for compatibility with propagation models to enable orbit prediction. Together with the traffic arriving from/to OGSs (B) and the assembly of traffic into chunks (C), this input enables the scheduling of optical paths, including routing and wavelength allocation. Four OGSs covering Europe are deployed [12], namely Maspalomas (27.76◦ ,15.59◦ ), Athens (37.98◦ ,23.72◦ ), Milan (45.4642◦ , 9.1900◦ ), and Bilbao (43.2630◦ , -2.9350◦ ). While the Grid+ topology connections across the LISLs are permanent, uplink and downlink connections are dynamically established based on line-of-sight and are periodically updated via a snapshot-based approach, with an average of 50 snapshots per simulation. To identify these hotspots, we use the Ookla and Human Settlement Proxy datasets, combining latency measurements with urban development indicators to capture current and future high-demand areas. TABLE II: Constellations configuration Constellation Sats Planes Inc. (deg) Altitude (km) Starlink Phase1 1584 72 53 550 Telesat 220 20 50.88 1325 Amazon Leo shell 1 784 28 33 590
B. Algorithms As illustrated in Fig. 2, a path computation component at the ground station (E) runs routing and wavelength reservation algorithms to select lambdas (e.g., from λ1 to λ4 ) in the optical up- and downlink based on WDM. The scheduler also needs an algorithmic solution to determine the best time to launch the transmission (F). A channel estimator (D), based on the channel model described below, uses information such as link distance and SNR from up-, down, and LISLs, and associated interferences to estimate the feasible transmission bitrate. It should be noted that a control channel is out-of-band (either RF or dedicated optical) and it contains information about resource reservation and scheduling algorithms, to be transmitted ahead of the corresponding data traffic (offset time). Control messages undergo optical-to-electrical (OE) conversion τOE , electrical processing, and electrical-to-optical (EO) conversion τEO at each satellite node and also control
the optical switch fabric for a proper switching configuration before the data is transmitted. As shown in Figure 2, λC , is used as the dedicated control wavelength, while wavelengths λ1 –λ4 are data plane; such separation is common in terrestrial optical networks. The performance of all-optical satellite networks is highly dependent on the interplay between optical switching fabric and the traffic chunk size, and presents a fundamental design tradeoff: while i) larger data chunks occupy network resources for longer periods and can increase contention, ii) smaller chunks may lead to poor resource utilization when the switching time is non-negligible, and E2E latency becomes dominated by switching times rather than data transmission time. The channel capacity (D) for each of the LISLs, as well as the up- and downlink, is estimated based on the following channel model. The up- and downlink data rates are expressed as a function of the signal-to-noise ratio (SNR) and the bandwidth B as [13]: C = B log2 (1 + SNR)
(1)
The SNR is determined by the received power PR and the system noise power N0 , and is given by SNR =
PR . N0
(2)
The received power is obtained from the link-budget formulation, based on [14], [15]: PR =
PT GT ηT GR ηR LFSO Latm Lfc Lp , Nch
(3)
where GT and GR denote the transmitter and receiver gains, respectively, while ηT and ηR represent their corresponding optical efficiencies. Furthermore, LFSO denotes the free-space optical path loss, Latm accounts for atmospheric effects, including scintillation-induced beam wander and turbulence, and Lp represents the pointing-error loss. The term Lfc denotes the additional coupling or channel-related loss considered in the link budget. Each loss component is briefly described below. During propagation from the OGS to the satellite, the signal is attenuated by several factors, starting with the free-space path loss (FSPL), which depends on the slant range R(ϑel ), elevation angle ϑel , and wavelength λ: LFSO =
λ 4πR(ϑel )
2 (4)
The signal also experiences atmospheric attenuation caused by scattering from particles along the propagation path. Using Kim’s model, this loss depends on the geometrical scattering attenuation coefficient αscatt (V, λ) and the Mie scattering defined based on the extinction ratio ρ and elevation angle. It is given by [14], [16]: Latm,att = αscatt (V, λ)R(ϑel ) + exp(−
ρ ) sin(ϑel )
(5)
Finally, in the uplink, atmospheric turbulence can cause scintillation-induced beam wander, leading to misalignment loss [15]. This loss is characterized using the Gamma-Gamma distribution, where α and β denote the small- and large-scale turbulence parameters, respectively [14]. The received optical intensity is denoted by I, while Γ(·) and K(·) represent the Gamma function and the modified Bessel function, respectively [14]. fGG (I) =
p 2(αβ)(α+β)/2 α+β −1 I 2 Kα−β (2 αβI) Γ(α)Γ(β)
(6)
For the uplink, the atmospheric loss is modeled as the combination of atmospheric attenuation and scintillation-induced beam-wander loss, i.e., Latm = Latm,att + Lsc
(7)
In contrast, for the downlink, only atmospheric attenuation is considered [15], yielding Latm = Latm,att
(8)
The pointing error losses for both LISLs and up- and downlink are calculated based on the Irradiance displacement model from [13]: 2
2
fhP E (h) = γ 2 A−γ hγ −1 , 0 with A0 = [erfc(v)]2 and v = √
0 < h ≤ A0 . √ √πwd , 2wz
γ =
(9) weq 2σp
and
πerfc(v) wz2eq = wz2 2v exp(−v 2 ) . In which ωd is the receiving aperture
radius, ωzeq is the equivalent beam width, and σp is the jitter standard deviation. Using the pointing loss obtained from Eq 9 the received power for the LISL can be defined as: PT PR = GT GR ηT ηR LF SO Lf c Lp Nch
delay, with the bottleneck link limiting the path’s effective throughput. Finally, switching delay τsw represents the reconfiguration time of the optical switching fabric. This parameter varies significantly with underlying hardware, ranging from milliseconds for MEMS to nanoseconds for SOAs. The subsequent analysis in this paper quantifies the impact of multiple devicedependent switching delays on overall system performance. Different optical switching paradigms can be adopted in alloptical satellite constellations, and selecting the best option is a challenge. While OPS-enabling technology and optical buffering are not mature enough, OCS, OBS, and their variations are compliant with all-optical switching, eliminate the need for complex optical buffering onboard satellites, and facilitate the SWaP requirements [2]. OCS improves reliability with the round-trip reservation, but has a penalty of increased E2E latency due to ACK messages. OBS drastically reduces the reservation period by sending data bursts right after the reservation message, but it might have reduced reliability when no ACK messages are used. In this work, a chunk represents an aggregated traffic unit transmitted after scheduling. Its size plays a key role in determining the trade-off between switching speed and E2E latency, which is central to the proposed chunkbased scheduling approach. To study this trade-off for various switching fabrics, we analyze the blocking ratio metric against E2E latency and energy metrics. The blocking ratio (BR) is defined as the percentage of chunks dropped due to resource unavailability or line-of-sight losses out of the total number of generated chunks, i.e., BR =
(10)
Thus, the LISL capacity can be obtained from Eq. 1, using the SNR defined in Eq. 2. C. Launching the Chunk When the data is ready to launch, a reservation packet (G), which contains information about the E2E path and the chosen wavelength, reserves the computed resources obtained from (E), and it is sent over λC . Afterwards, the transmission of the traffic chunk (H) is conducted. The overall E2E latency in such all-optical satellite constellations includes four primary components: processing, propagation, transmission, and switching delays. Processing delay τproc is defined as the time required to process and execute control-plane signaling, which is typically transmitted in advance to pre-configure the optical switching fabric for incoming data chunks. The propagation delay τprop is determined by the speed of light across uplink, downlink, and LISLs, and scales with orbital altitude, constellation density, and inter-satellite distances. The transmission delay τtr corresponds to the time required to inject a data payload onto the physical medium. Thus, larger payloads or lower-bandwidth links contribute to this
NDropped ∗ 100 NTotal
(11)
As previously mentioned, E2E latency (L) is a parameter constrained in modern constellations. To calculate this value in our systems, we sum up the following delay components on optical paths, i.e., transmission delay τtr at the bottleneck link, propagation delay τprop , and onboard satellite delays, including processing and optical switching, i.e., X X L = τtr + τprop + (τproc + τswitch ) (12) Hops
Sats
Since optical switches contribute to the energy budget onboard a satellite, we also study the related energy efficiency (EE), defined as the total number of successfully delivered bits divided by the total power consumption of optical switching hardware, normalized by the losses due to blocking, i.e., P EE (bits/W) =
Bitsreceived chunks P sats Pswitch
∗ (1 − BR)
(13)
The presented BR metric in Eq. 11 allows for analyzing how the different chunk sizes affect the resource availability and allocation, as bigger data chunks imply higher transmission delays, longer channel reservation, and increased competition for free resources. Larger transmission delays associated with
10 5 2 1
100
0 1 10 50 100 200 300 350 400 500 550 600 1000
Polatis GLSUN AGILTRON SOA
10 5 2 1
100
Blocking ratio (%)
Polatis GLSUN AGILTRON SOA
Blocking ratio (%)
Blocking ratio (%)
100
0 1 10 50 100 200 300 350 400 500 550 600 1000
Chunk size (MB)
10 5 2 1
0 1 10 50 100 200 300 350 400 500 550 600 1000
Chunk size (MB)
(a) Telesat
Polatis GLSUN AGILTRON SOA
Chunk size (MB)
(b) Amazon Leo shell 1
(c) Starlink phase 1
80
Propagation Telesat
Transmission Amazon Leo
Switching Starlink
60 40 20 0
1 10 50 100 200 300 350 400 500 550 6001000
Chunk size (MB)
120
End-to-end Latency (ms)
End-to-end Latency (ms)
100
100
Propagation Telesat
Transmission Amazon Leo
Switching Starlink
150
Propagation Telesat
Transmission Amazon Leo
Switching Starlink
125
80
100
60 40 20 0
End-to-end Latency (ms)
Fig. 3: Blocking ratio for different chunk sizes vs. switching fabrics for various constellations
1 10 50 100 200 300 350 400 500 550 6001000
(a) SOA and AGILTRON
Chunk size (MB)
(b) GLSUN
75 50 25 0
1 10 50 100 200 300 350 400 500 550 6001000
Chunk size (MB)
(c) POLATIS
Fig. 4: E2E latency breakdown for different chunk sizes, switching technologies, and constellations.
large chunks are expected to result in increased overall E2E latency (Eq. 12). Finally, EE measures the efficiency of bits transmitted by the power unit, according to the different power consumption for the different switching devices. IV. S IMULATION SETUP AND EVALUATION We now analyze the results based on the simulation. The traffic assembly and chunk forwarding is similar to what is known for terrestrial OBS, but in its approach also assembles fast circuit switching, whereby circuit holding time can be approximated to match the chunk size based on channel capacity. The traffic assembly follows a maximum traffic chunk (burst) size threshold, which triggers the scheduler once achieved. We do not make assumptions on control protocol, whereby different reservation protocols (e.g., 1-way, 2-way) and introduction of offset times (τof f in Fig. 2) can be taken into the overall E2E latency, which is out of scope here. Event-driven simulations are carried out using OMNeT++ v6.3.0. The OS3 library is employed to generate satellite constellations shown in Table II, including their propagation models based on TLE data obtained via [17]. The newly developed traffic forwarding and routing library is integrated with OS3, and it also includes features such as multi-channel WDM links, an optical channel estimator, wavelength assignment, traffic assembly, scheduling, routing, and Grid+ LISLs. Wavelength assignment is simple and is performed using a first-fit strategy, selecting the earliest available wavelength. Routing then follows the K-shortest paths algorithm with k = 5, yielding average satellite hop counts of 1.46, 2.56, and 2.85 for Telesat, Amazon Leo, and Starlink, respectively. Traffic is modeled with an Erlang distribution (k = 1, rate λ),
i.e., a Poisson arrival process. To keep loads comparable across traffic chunk sizes, λ is set based on the bottleneck channel capacity and the corresponding transmission delay for each size. The remaining simulation parameters are summarized as follows. The OGS is assumed to have an aperture diameter of 0.8 m, corresponding to a gain of 125 dBi. The satellite transmit power is set to 0.7 W, with a satellite antenna gain of 120 dBi. Clear-sky weather conditions are assumed at the OGS location; accordingly, the visibility is set to 60 km. The optical wavelength is set to 1550 nm. Fig. 3 presents the chunk blocking rate (Eq. 11) for various chunk sizes (varying from 1MB to 1GB) for Telesat, Amazon Leo, and Starlink constellations and across different switches. A clear trend is noticed for both ms-switches (POLATIS and GLSUN), which yield high losses for small chunks. For instance, POLATIS yields losses greater than 2% for chunk sizes up to 400MB in all constellations due to the slower switching mechanism, which is the dominant factor in the E2E latency shown in Fig. 4c. The same behavior occurs for GLSUN, up to a chunk size of 50 MB in Telesat (Fig. 3a) and 200 MB in Starlink and Amazon LEO. This early convergence with Telesat happens due to its broader coverage and the lower number of hops crossed, reducing the switching time and making propagation the dominant delay for smaller chunk sizes in Fig.4b. On the other hand, µs- and ns-switches show the same sub 1% traffic loss across different chunk sizes, due to negligible switching delay, as Fig. 4a shows. We consider E2E latency range between 30 and 60 ms to represent the latency range experienced by current commercial satellite networks [18], and establish this as a constraint and target in performance evaluation of various constellations.
Fig. 4 illustrates the latency breakdown for different switching fabrics. The Telesat constellation has the highest altitude, thus the highest propagation delay, followed by Amazon LEO and Starlink, respectively. For SOA and AGILTRON (Fig. 4a), the maximum feasible traffic chunk sizes under the 60ms threshold are 600 MB (Amazon LEO and Starlink) and 500 MB (Telesat). Chunk sizes of 600 MB cause a 0.83% loss rate in Starlink, and 0.68% in Amazon Leo. Telesat, in turn, suffers from 0.98% loss with 500 MB chunks. Therefore, Amazon Leo with 600 MB demonstrates 0.15% lower loss and 0.21 ms higher E2E latency than Starlink. For the ms-scale switches, GLSUN (Fig. 4b) illustrates a maximum feasible chunk size of 350 MB for Telesat and Amazon LEO, and 300 MB for Starlink. The increased switching delay reduces the time available for transmission, thus reducing the size of the traffic chunk for feasible transmission. Finally, Polatis switches (Fig. 4c) only perform under the E2E latency threshold in Telesat for chunks with 100 MB size, due to the high added switching delay, while for the other constellations, chunk sizes of 1 MB already exceed the 60 ms E2E delay threshold. For POLATIS switches, the transmission delay only becomes dominant for bigger chunks (1 GB). Reduced chunk sizes are dominated by switching delay and have often optical channels in an idle state. Based on the analysis, the maximum feasible chunk sizes for each switch and constellation were selected and compared in terms of their energy efficiency, losses, and E2E latency, as illustrated in Table III. Overall, SOA achieves the best performance across all evaluated metrics. Among the other solutions, GLSUN deployed in Telesat has the highest energy efficiency for 350 MB chunk sizes. The results show that the choice of appropriate chunk sizes varies alongside many aspects of the scenario, such as the constellation layout and the adopted switching fabric, and a trade-off evaluation enables fine-grained adjustment of the appropriate traffic chunk size. TABLE III: Energy efficiency evaluation Arch. SOA
Const. Telesat Amazon LEO Starlink AGIL. Telesat Amazon LEO Starlink GLSUN Telesat Amazon LEO Starlink POLATIS Telesat
Traffic BR (%) E2E Latency (ms) EE (Gbits/W) 500MB 0.98 59.91 4.55 600MB 0.68 60.68 3.22 600MB 0.83 60.47 2.89 500MB 0.98 59.91 0.26 600MB 0.68 60.68 0.18 600MB 0.82 60.47 0.16 350MB 1.16 58.04 1.48 350MB 1.01 60.65 0.87 300MB 1.33 58.70 0.66 100MB 2.4 59.03 0.11
V. C ONCLUSION We investigated the tradeoff between switching fabric design and traffic chunk sizing for modern constellations with all-optical switching. The results show that ms-scale switches incur high losses for small chunks and require large chunks to meet the 30-60 required latency, still experiencing >1% loss, thus severely restricting feasible operation. In contrast, µs- and ns-switches achieve a feasible chunk size of 500MB for Telesat and 600 MB for Amazon Leo and Starlink, while maintaining
the 60 ms latency threshold and loss ratio <1%. Considering energy efficiency, ms-scale switches can, in some cases, outperform certain fast-switching alternatives, indicating that the optimal choice of traffic profile and constellation is ultimately driven by system priorities as different configurations involve trade-offs across loss, latency, and energy efficiency. ACKNOWLEDGMENT This work was partly funded by the European Space Agency (ESA) in the framework of the project NEXON (activity number 1000042074). R EFERENCES [1] S. Ma and et. al., “Network characteristics of leo satellite constellations: A starlink-based measurement from end users,” in IEEE INFOCOM 2023 - IEEE Conf. on Computer Communications, 2023, pp. 1–10. [2] F. Li and et. al., “A review on all-optical switching in intersatellite laser communication,” Advanced Devices & Instrumentation, vol. 6, p. 0097, 2025. [3] Y. Wang and et. al., “Space-based optical burst switching assembly algorithm based on qos adaption,” in International Conference on Communication Software and Networks (ICCSN). IEEE, 2019. [4] C. Wang and et. al., “Osbn: architecture and control mechanism of optical switched satellite backbone network,” Photonic Network Communications, vol. 43, p. 165–176, 2022. [5] Z. Li and et. al., “100 gb/s per channel dynamic adaptive granularityaware onboard optical switching,” IEEE Internet of Things Journal, vol. 13, no. 1, 2026. [6] Q. Cheng, S. Rumley, M. Bahadori, and K. Bergman, “Photonic switching in high performance datacenters [invited],” Optics Express, vol. 26, no. 12, 2018. [7] H. Zhai and et. al., “Design and implementation of the hardware platform of satellite optical switching node,” in 2021 19th International Conference on Optical Communications and Networks (ICOCN). IEEE, 2021, pp. 01–03. [8] Guilin GLsun Science and Tech Group Co., LTD, “4x4 cascade optical switch,” 2025. [Online]. Available: https://www.glsun.com/product-v594x4-cascade-optical-switch.html [9] Agiltron Inc., “2×4, 4×4 high speed fiber optical switch system,” 2024. [Online]. Available: https://agiltron.com/product/4x4-high-speedoptical-switch-nanospeed/ [10] HUBER+SUHNER Polatis, “Polatis series 6000s optical matrix switch,” 2024. [Online]. Available: https://www.hubersuhner.com/en/shop/product/other-systems/opticalswitches/rack-mount-circuit-switches/85222533/polatis-6000s-opticalcircuit-switch [11] D. W. Feyisa and et. al., “Compact 8×8 soa-based optical wdm space switch in generic inp technology,” Journal of Lightwave Technology, vol. 40, no. 19, 2022. [12] O. Farley and J. Osborn, “Optimising optical ground station locations for satellite communications through atmospheric turbulence with adaptive optics mitigation,” in ECOC 2024; 50th European Conference on Optical Communication, 2024, pp. 398–401. [13] B. Shang, , and et. al., “Channel modeling and rate analysis of optical inter-satellite link (oisl),” IEEE Transactions on Vehicular Technology, vol. 74, no. 7, pp. 11 650–11 655, 2025. [14] V. Spirito and et. al., “E2e physical layer and link analysis for highthroughput satellite optical communication,” IEEE Journal on Selected Areas in Communications, vol. 43, no. 5, pp. 1660–1675, 2025. [15] C. Cantore and et. al., “Link budget analysis of bi-directional leo and geo optical feeder links advancing the beam wander model’s accuracy,” Scientific Reports, vol. 14, no. 1, p. 8579, 2024. [16] E. Erdogan and et. al., “Site diversity in downlink optical satellite networks through ground station selection,” IEEE Access, vol. 9, pp. 31 179–31 190, 2021. [17] “Bulk TLE Generator — tle-generator.starlitter.info,” https://tlegenerator.starlitter.info/, [Accessed 18-04-2026]. [18] M. Dano. (2026, Feb.) 2025 global satellite broadband performance report. Ookla. Accessed: 2026-04-11. [Online]. Available: https://www.ookla.com/articles/2025-global-satellite-broadbandperformance-report