PaperlayerView exact revision on arXiv

Source-linked AI summary

Exploring the Initial Performance of NB-IoT NTN over GEO: Measurement and Analysis

Xiong Wang, Yumin Du, Bin Gu, Zheng Lin, Xiaoyang Li, Yi Gong, Wei Gong, Linghe Kong

arXiv:2609.06280v1cs.NI

TL;DR

Operational behavior of NB-IoT NTN over GEO remains insufficiently documented, despite its role in extending IoT connectivity by satellite. The paper measures two commercial providers and analyzes delay and energy; its results identify access-related costs and motivate designs to reduce them.

  • Problem

    Large-scale, reproducible accounts of operational NB-IoT hardware and satellite deployments, including divergence between models and measured behavior, remain scarce.

    Source: §7 · p. 12

  • Method

    The authors conduct a more-than-six-month end-to-end measurement study of Skylo and Tiantong, focusing on delay and energy, and develop models, breakdown analyses, and trace-driven design evaluations.

    Source: §1 · pp. 1–2

  • Results

    Energy consumption increases linearly with E2E delay and is mainly incurred by MSG3 retransmissions and NAS overhead, while data transmission occupies only 5%; access delay also rises linearly with concurrent terminals.

    Source: §1 · p. 2

  • Takeaways & Limitations

    The measurements and proposed designs provide a basis for future implementation and deployment of IoT-NTN over GEO.

    Source: §8 · p. 12

    The study uses ten ground terminals for concurrency measurements and omits link reliability under different weather conditions and antenna types.

    Source: §6 · p. 11

Abstract

from arXiv · show

With the standardization of Non-Terrestrial Networks (NTN) to provide direct satellite connectivity to massive, low-power Internet of Things (IoT) devices, 3GPP IoT-NTN bridges the worlds of cellular and satellite communications. While holding great potential for global connectivity with IoT devices, there exist several concerns about the system performance of IoT-NTN over Geostationary Earth Orbit (GEO), which covers multiple dimensions such as end-to-end delay and energy consumption considering the ultra-long transmission distance from ground IoT terminals to GEO satellite. To answer these concerns, we have conducted the first comprehensive and in depth measurement for NB-IoT NTN over GEO. Based on real NB-IoT NTN testbeds including both Skylo and Tiantong, measurements covering more than six months confirm that the available implementation of NB-IoT NTN remains in the initial stage. Amplified by ultra-long Round-Trip Time (RTT) between ground IoT terminals and GEO satellite, there exists plenty of time and energy consumption during the access process. Interestingly, it also reveals that as an energy-saving mechanism, Power Saving Mode (PSM) fundamentally reshapes NB-IoT NTN traffic into a bursty and access-driven communication pattern and thus has a considerable influence on the end-to-end delay and energy consumption. Finally, we propose corresponding optimization schemes to reduce the delay and energy, which lays a good foundation for the implementation of NB-IoT NTN in the near future.

1 Introduction

The paper introduces a first long-term, two-provider measurement study of NB-IoT NTN over GEO, examining end-to-end delay and energy consumption. Its findings identify access-related inefficiencies and motivate protocol designs to reduce delay and energy.

1 Introduction

  • The study measures Skylo and Tiantong for more than six months, covering E2E delay and energy consumption with JMD 47 ground terminals.The measurements reveal mismatches between standard design and real satellite deployments.Source: §1 · p. 1
  • Packet payload size affects GEO E2E delay because NB-IoT NTN uplink transmissions have limited TBS, although the effect is almost negligible in terrestrial networks.The introduction characterizes E2E delay as control-plane dominated rather than traffic-dependent.Source: §1 · p. 2
  • Skylo uses more suitable settings, including a small TBS configuration, than Tiantong; the authors infer that NB-IoT NTN over GEO is optimized for small packets.Small packets can be delivered with MSG3 or NAS, given the ultra-long RTT.Source: §1 · p. 2
  • Access delay increases linearly with the number of concurrently transmitting ground terminals, creating long-tail latency.Short packet intervals intensify the effect of MAC access and HARQ/NPUSCH retries under ultra-long GEO RTT.Source: §1 · p. 2
  • Energy consumption increases linearly with E2E delay, driven mainly by MSG3 retransmissions and NAS overhead, while data transmission accounts for only 5%.The authors also report that PSM reshapes traffic into a bursty, access-driven pattern.Source: §1 · p. 2
  • The paper models delay and energy, breaks down bottlenecks, and proposes designs evaluated with trace-driven simulations.Its stated contributions also include analysis of how PSM, packet interval, and payload size shape delay.Source: §1 · p. 2

2 Background and Motivation

GEO IoT-NTN reuses cellular procedures over a satellite link whose propagation delay creates challenges for protocol operation. Although standards and models address NTN, real-world GEO behavior remains insufficiently understood, motivating empirical measurement.

2.1 Background

  • A GEO satellite relays IoT-device traffic to a gateway and terrestrial core; its distance yields RTT exceeding 500 ms, affecting random access, scheduling, and retransmissions.The satellite is approximately 35,786 km above Earth, with one-way propagation delay around 120–140 ms.Source: §2.1 · p. 3
  • The conventional uplink access procedure uses contention-based MSG1–MSG4, then establishes RRC, sends data, listens for downlink traffic, and releases the connection.If MSG4 does not arrive before the contention resolution timer expires, the terminal re-initiates random access.Source: §2.1 · p. 3
  • 3GPP reuses cellular mechanisms with incremental NTN adaptations, but GEO's much larger delay raises questions about terrestrial-designed protocol performance.Adaptations include extended timing advance, modified random access, relaxed latency constraints, and optional transparent or regenerative payloads.Source: §2.1 · p. 3

2.2 Motivation

  • Despite standardization and industrial interest, real-world 3GPP IoT-NTN behavior over GEO remains largely unexplored.Available measurement work includes simulations, but GEO link characteristics remain insufficiently understood through simulations or analytical models alone.Source: §2.2 · p. 3
  • GEO RTT typically exceeds 500 ms and affects core procedures, while the effectiveness of specification parameters based on analytical assumptions and laboratory emulation remains uncertain in deployments.Low-cost devices also face long timing advance, oscillator drift, asymmetric paths, and gateway queuing, leaving operational reliability, latency, and energy efficiency unknown.Source: §2.2 · pp. 3–4
  • These uncertainties motivate real-world GEO experiments to measure delay, energy, and reliability and inform future NTN design and deployment.The stated goal is to expose behaviors not visible in simulation and validate or challenge standard assumptions.Source: §2.2 · p. 4

3 End-to-end Delay

The section measures end-to-end delay across payload sizes, reporting intervals, and concurrent access in GEO NB-IoT NTN. Results show that delay reflects both fixed propagation and control overhead and access contention.

3 End-to-end Delay

  • End-to-end delay is measured with ping from a ground terminal through the GEO satellite, ground station, and core network to a Hong Kong cloud server.Each configuration is measured 30 times across Skylo and Tiantong.Source: §3 · p. 4

3.1 Under different payload size

  • In Tiantong, idle-state delay can be lower for small packets because MSG3 may carry data directly, while linked-state signaling is steadier; large packets narrow or reverse the difference.The direct MSG3 path is prone to resource conflicts and delay fluctuations.Source: §3.1 · pp. 4–5
  • Skylo and Tiantong show system-specific payload thresholds of about 32 and 128 bytes, respectively; above the threshold, serialized uplink fragments drive near-linear delay growth.Skylo delay is centered around 9 seconds below 32 bytes, then reaches 11–12 seconds at 64 bytes and over 15 seconds at 128 bytes.Source: §3.1 · pp. 4–5

3.2 Under Different Traffic Interval

  • At a 2-second interval, delay can exceed 17 seconds, while 10-second pacing is flatter; larger payloads, especially at short intervals, raise delay variability and can incur packet loss.For 16–32-byte packets, delay remains relatively stable across intervals; 64–128-byte packets show greater delay and dispersion, particularly at 2 seconds.Source: §3.2 · pp. 5–6

3.3.1 Concurrency=1

  • Even with one device, GEO NB-IoT NTN has a baseline delay of multiple seconds, primarily from satellite round-trip propagation and remote core-network processing.The measurement uses Skylo terminals, with concurrency levels representing one to ten simultaneously transmitting devices.Source: §3.3.1 · p. 6, §3.3 · p. 6

3.3.2 Concurrency>1

  • As concurrency rises, delay curves shift upward and skew; the study reports superlinear delay growth and tails exceeding 30 seconds for high-index devices.The authors attribute this to long RTT combined with contention-based random access, sequential scheduling, collisions, delayed grants, and retransmissions.Source: §3.3.2 · pp. 6–7, Figure 7 · p. 6
  • Beyond six concurrent devices, delay distributions widen sharply, with medians around 10–15 seconds and upper whiskers beyond 35 seconds.At concurrency of one to two devices, delay is concentrated around 8–9 seconds.Source: §3.3.2 · p. 7, Figure 8 · p. 7
  • For delay-sensitive or synchronized applications, median latency alone misses the constraint imposed by worst-case access times; the authors identify staggering, concurrency control, or prioritization as possible mitigations.The concurrency measurements use up to ten NB-IoT nodes transmitting simultaneously over Skylo.Source: §3.3 · p. 6, §3.3.2 · p. 7

4 Energy Consumption vs Delay

The section links energy use to end-to-end delay and examines how connection procedures, traffic intervals, and PSM shape that cost. Its measurements show that signaling and access can outweigh data transmission, with GEO delay making traffic and access behavior consequential for both energy and latency.

4 Energy Consumption vs Delay

  • The dominant energy term scales linearly with end-to-end delay, so reducing access repetitions and connection time minimizes both latency and energy.The paper describes this as an energy–latency tradeoff.Source: §4 · p. 7

4.1 Power consumption using different transmission schemes

  • The traditional UP-CIoT idle connection takes 8–9 seconds; CP-CIoT reduces delay to around 6 seconds, while UP-CIoT Hot Start slightly outperforms it in latency and energy.Hot Start retains network context to skip some authentication and bearer-establishment steps; it is recommended for intermittent or short-cycle transmissions when signaling does not time out.Source: Figure 9 · p. 8, §4.1 · p. 8
  • MSG3 retransmissions account for 56% of uplink energy, NAS signaling for 28%, and data transmission for 5%.The energy breakdown highlights protocol operation, rather than payload delivery, as the larger cost.Source: Figure 10 · p. 8, §4.1 · p. 8

4.2 Power consumption trace with different traffic interval

  • Sending 10 packets at 2-, 5-, and 10-second intervals consumes 0.0101, 0.0199, and 0.0262 Wh, respectively; the 10-second interval uses almost twice the energy of the 2-second interval.At 2 seconds the terminal remains connected; at 10 seconds it sleeps or enters RRC idle and must repeat access signaling.Source: Figure 11 · p. 9, §4.2 · p. 9
  • Around a 5-second interval, jitter near the likely RRC inactivity-timer threshold can trigger inconsistent connection retention or re-access.The authors recommend recalibrating and extending the inactivity timer for GEO RTT or avoiding application intervals that trigger this jitter.Source: §4.2 · p. 9

4.3 PSM

  • PSM creates stepped energy use and bursty, access-driven traffic; its interaction with long RTT amplifies access latency and variability.The authors recommend aligning PSM configuration with application traffic patterns.Source: §4.3 · p. 9, Figure 12 · p. 10

5 Implications for NB-IoT NTN Design

The design implications focus on reducing access competition before access and MSG3 retransmissions afterward, alongside adaptive controls for PSM timers and access barring. Simulations report improved energy, delivery, collision, and retransmission outcomes for the proposed schemes.

5 Implications for NB-IoT NTN Design

  • Before access, minimize access competition; after access, minimize MSG3 retransmissions for resource requests.Source: §5 · p. 10
  • Under moderate channel conditions, the adaptive scheme consumes 44% less power than r32 and 83% less than r128, respectively. Under worst channel conditions, it consumes 44% less energy than r128 for MSG3 reception.Under good channel conditions, RACH success rates are 89% for the adaptive scheme and 96% for r4.Source: §5 · p. 10
  • The adaptive scheme selects MSG3 retransmission numbers of 4, 17, and 61 across the three ECL levels, compared with 4, 32, and 128 for the available scheme.The scheme estimates MSG1 reception power to determine the required MSG3 retransmission number.Source: §5 · p. 10, Figure 15 · p. 11
  • 85.2% DDR with 1683 J average energy: MIDA improves on fixed-timer baselines while adapting T3324 to delivery outcomes.A lost packet doubles T3324, while three consecutive successful deliveries shrink it by a fixed step.Source: §5 · pp. 10–11
  • MIDA approaches the oracle’s energy-DDR Pareto frontier within 17% in energy and 15% in DDR.The oracle achieves 100% DDR at an energy cost of 1433 J.Source: §5 · p. 11
  • At more than 100 ground terminals, dynamic ACB holds collision rate at 10%, while static ACB rises from 13% to 23% and ALOHA from 23% to 34%.The proposed barring factor adjusts according to satellite-side per-second collision rate.Source: §5 · p. 11
  • As terminals increase from 20 to 100, the ACB barrier factor rises from 0.10 to 0.68, reducing collision rate from 34.3% to 11.2%.The ACB-aware scheme also brings a higher successful access probability than ALOHA.Source: §5 · p. 11

6 Discussion and Limitations

The discussion identifies limited concurrency and incomplete measurement coverage as boundaries on the study. Its timer-setting figures present delivery, energy, and combined trade-off comparisons.

6 Discussion and Limitations

  • The concurrency experiment used ten ground terminals, while the authors note that actual scenarios may involve hundreds or thousands and that delay distributions widen as concurrency grows.Source: §6 · p. 11
  • Measurements cover E2E delay and power consumption but omit link reliability across weather conditions and different antenna types.The authors leave those measurements for future work.Source: §6 · p. 11
  • The timer-setting figures compare Downlink Delivery Ratio, energy consumption, and the Energy-DDR Pareto frontier.Source: Figure 16 · p. 12, Figure 17 · p. 12, Figure 18 · p. 12

7 Related Work

Related work spans standardization, analytical and simulation studies, and prototype efforts, but the paper identifies a shortage of systematic long-term operational accounts. It positions its work against that gap.

7 Related Work

  • Prior work covers 3GPP NTN architecture and scenarios, plus NB-IoT satellite adaptations and analytical or simulation studies of access, synchronization, coverage, and power-latency trade-offs.Source: §7 · p. 12
  • Prototype and testbed efforts report practical issues including RF constraints, antenna and pointing trade-offs, Doppler compensation, and intermittent contacts.Source: §7 · p. 12
  • The literature lacks large-scale, reproducible accounts of long-term operational telemetry, integration decisions, and divergence between model predictions and measured behavior.Source: §7 · p. 12

8 Conclusion

The conclusion presents measurements of initial IoT-NTN performance over GEO and simple designs addressing measured inefficiencies. It describes these contributions as a foundation for future implementation and deployment.

8 Conclusion

  • The paper measures and investigates initial IoT-NTN network performance over GEO.Source: §8 · p. 12
  • It proposes simple, effective designs to address inefficiencies identified through measurement and analysis.Source: §8 · p. 12
  • The authors describe the work as a foundation for future IoT-NTN implementation and deployment over GEO.Source: §8 · p. 12
Loading 2609.06280v1…