TSVWG Group S. Yue
Internet-Draft China Mobile
Intended status: Informational l. Izhikevich
Expires: 7 January 2026 University of California
T. Tsou
Tiktok
6 July 2025
The Use Cases for Optimizing Transport Layer Protocols in LEO and MEO
Satellite Networks
draft-yue-tsvwg-leo-transport-use-cases-00
Abstract
This document introduces the use cases for the performance of
existing transport protocols in LEO and MEO satellite networks. The
use cases identify and demonstrate the current challenges faced by
transport layer protocols in these environments.
Status of This Memo
This Internet-Draft is submitted in full conformance with the
provisions of BCP 78 and BCP 79.
Internet-Drafts are working documents of the Internet Engineering
Task Force (IETF). Note that other groups may also distribute
working documents as Internet-Drafts. The list of current Internet-
Drafts is at https://datatracker.ietf.org/drafts/current/.
Internet-Drafts are draft documents valid for a maximum of six months
and may be updated, replaced, or obsoleted by other documents at any
time. It is inappropriate to use Internet-Drafts as reference
material or to cite them other than as "work in progress."
This Internet-Draft will expire on 7 January 2026.
Copyright Notice
Copyright (c) 2025 IETF Trust and the persons identified as the
document authors. All rights reserved.
Yue, et al. Expires 7 January 2026 [Page 1]
Internet-Draft The Use Cases for LEO and MEO July 2025
This document is subject to BCP 78 and the IETF Trust's Legal
Provisions Relating to IETF Documents (https://trustee.ietf.org/
license-info) in effect on the date of publication of this document.
Please review these documents carefully, as they describe your rights
and restrictions with respect to this document. Code Components
extracted from this document must include Revised BSD License text as
described in Section 4.e of the Trust Legal Provisions and are
provided without warranty as described in the Revised BSD License.
Table of Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 2
2. Conventions and Definitions . . . . . . . . . . . . . . . . . 3
3. Use Cases . . . . . . . . . . . . . . . . . . . . . . . . . . 3
3.1. Video Streaming Service over LEO . . . . . . . . . . . . 3
3.2. 3GPP UE-SAT-UE Communication for 5G VN Group Service . . 4
3.3. LEO Satellite Networks vs. Cellular Network . . . . . . . 6
4. Security Considerations . . . . . . . . . . . . . . . . . . . 7
5. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 7
6. References . . . . . . . . . . . . . . . . . . . . . . . . . 7
6.1. Informative References . . . . . . . . . . . . . . . . . 7
6.2. Normative References . . . . . . . . . . . . . . . . . . 8
Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 8
1. Introduction
With the rapid advancement of Low Earth Orbit (LEO) and Medium Earth
Orbit (MEO) satellite networks, these systems have become
increasingly important for providing global connectivity, especially
in remote and underserved areas. However, compared to traditional
terrestrial networks, LEO and MEO satellite networks face significant
challenges due to their unique characteristics, such as frequent
handovers, highly variable latencies, and periodic connection drops.
These extreme network dynamics pose substantial difficulties for
existing transport layer protocols, which are primarily designed for
more stable and predictable network environments.
Existing transport protocols, such as TCP, fail to perform optimally
in LEO and MEO satellite networks. For example, TCP's congestion
control mechanisms, which rely on packet loss as a primary signal for
congestion, can misinterpret the packet loss by handovers in
satellite networks as a sign of network congestion, leading to
unnecessary throughput reductions. Transport layer protocol
requirement for LEO satellite was introduced at
[I-D.yang-tsvwg-leo-transport-req] .
Yue, et al. Expires 7 January 2026 [Page 2]
Internet-Draft The Use Cases for LEO and MEO July 2025
This document introduces the use cases for the performance of
existing transport protocols in LEO and MEO satellite networks. The
use cases identify and demonstrate the current challenges faced by
transport layer protocols in these environments.
2. Conventions and Definitions
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
"SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and
"OPTIONAL" in this document are to be interpreted as described in BCP
14 [RFC2119] [RFC8174] when, and only when, they appear in all
capitals, as shown here.
3. Use Cases
3.1. Video Streaming Service over LEO
Netflix, as a major video streaming service, analyzed data over one
million Starlink users across 85 countries over two years in
[Izhikevich2024] . Starlink dominates the LEO satellite network
market as the service provider with near 6,000 satellites and over
2.6 million users. The analysis reveals that while Starlink’s
overall video quality matches traditional terrestrial networks,
inherent LEO network variability, including throughput fluctuations
and packet loss, causes a 60% rise in bitrate switches and a 200%
increase in rebuffering events.
Furthermore, the study highlights that Starlink exhibits lower
throughput and higher variance compared to non-LEO terrestrial
networks. For instance, Starlink's throughput is typically about 50%
of that offered by top 10 ISPs and falls below 20 Mb/s. Over 90% of
bitrate switches occur at throughputs below 20 Mb/s. Additionally,
Starlink's throughput recovery is slower: while non-LEO networks
recover to above 10 Mb/s in about 5 seconds, Starlink takes
approximately 15 seconds.This delay coincides with Starlink’s static
route reconfiguration intervals between satellites and ground
stations.
To investigate the causes of low throughput, Netflix deployed A/B
tests on over 1 million users. The study identified that Starlink's
high packet round trip times (RTTs) and increased retransmit rates
contribute to low TCP congestion windows, resulting in lower
throughput. Nearly two times longer RTTs, double the retransmit
rate, and smaller congestion windows compared to other networks were
observed to confirm that. The higher latency is attributed by longer
routing paths. The increase in Starlink retransmits is likely
influenced by random loss on satellite links; the use of Active Queue
Management on the Starlink WiFi router which minimizes latency at the
Yue, et al. Expires 7 January 2026 [Page 3]
Internet-Draft The Use Cases for LEO and MEO July 2025
expense of packet loss; and potential packet reordering which may
lead to duplicate acknowledgments and spurious retransmits.
Increases in RTTs and retransmit rates can cause loss-based
congestion controllers to take a long time to grow the TCP congestion
window. In addition, loss-based congestion controllers assume that
any packet loss is due to congestion in the network, and therefore,
back off when seeing indications of packet loss. Congestion windows
for Netflix sessions over Starlink paths being lower than over other
networks. With lower congestion windows, Starlink sessions are more
likely to underutilize the true available bandwidth for a single TCP
connection, and therefore experience lower throughput.
To achieve higher throughput, Netflix approximated the use of
multiple connections with MulTCP and created a modified version of
New Reno: it grows the congestion window by three bytes for every
acknowledged byte, and when it detects packet loss, it reduces the
congestion window as if only one of the three emulated flows had seen
a loss event. However, a 30%-40% throughput increase came at the
cost of increased retransmit rates by 300% and latency by 77%.
Further solution is needed to better handle the dynamic
characteristics of LEO networks and improve transmission performance.
3.2. 3GPP UE-SAT-UE Communication for 5G VN Group Service
3GPP is a Standard Development Organization or SDO that develops and
maintains technical spec. for mobile networks. It is currently
working on the final 5G-A spec. and simultaneously exploring various
kinds of domains for the coming 6G evolution.
3GPP has, in different releases, (near-) completed a few satellite-
related working lis, including Satellite-access, Satellite-backhaul,
Satellite regenerative forwarding mode, etc [TS.23.501]. As of now,
for satellite operators who are operating 5G network with satellite
access globally, the 3GPP plenary has concluded the requirements of
supporting UE-satellite-UE communication for 5G VN group service
[_GPP.SP-250400] . The 3GPP UE-Sat-UE service indicates the
communication, between UEs under the coverage of one or more NGSO
satellites, uses satellite access without the inter-UE traffic
transiting thru the ground segment.
As per the 3GPP [TS.23.501] , the 5G VN Group communication service
is comprised of two traffic forwarding types, i.e.,
* local-switch type: for which the UL/DL traffic is locally
forwarded by a single (on-board satellite) UPF (functioning like a
router) if this UPF is the ‘anchor’ for all the participating UEs
of the same 5G VN group;
Yue, et al. Expires 7 January 2026 [Page 4]
Internet-Draft The Use Cases for LEO and MEO July 2025
* inter-UPF type: for which the UL/DL traffic for the 5G VN group
communication is forwarded between/among ‘anchors’ of different
participating UEs.
Here, both the local-switch and the inter-UPF forwarding types
involve the impacts & implications as introduced by inter-satellite
links or ISLs. Note that, while the 3GPP simplifies the satellite
network architecture by assuming the existence of ISL(s) have no or
insignificant impacts on the 5G VN group service, we IETF must
consider various types of implications associated with the dynamic
satellite constellation networks, e.g.,
* High latency: thanks to potential long-distance satellite paths.
* high loss: harsh environment in outer space impairing the
transmission quality, e.g., radiation, solar storm, etc.
* Path instability: thanks to extremely dynamic topology of
satellite network, leading to frequent ISLs break-up &
reconnection between satellites, satellite handover, etc.
* Path switchover & multi-path impacts: related to equipment
features and network design, e.g., frequent beam scanning of
(phased array) antennas, peering changes between/among neighboring
satellites, UE mobility, etc.
Compared to the deployment over terrestrial networks (or TNs), the
traditional transport-level congestion control mechanism, e.g., TCP
CUBIC, etc., would perform much worse due to the above adversarial
factors. The APNIC blog [APNIC-blog] , based on investigational
survey, suggests it be possible to adjust the traditional TCP CUBIC
by conducting a lost packet repair using Selective Acknowledgement
(SACK) [RFC2883] if a packet loss event might occur at the time of a
scheduled satellite handover or path switchover. Remember that SACK
is a big reason why modern TCP protocols perform well over mobile
services. Fortunately, the 3GPP-based UE-Sat-UE 5G VN service does
own the capability of knowing in advance satellite scheduled changes.
In 5G network [TS.23.501] , the 5GC NF AMF can get the satellite
ephemeris information via different schemes and provide the satellite
footprints to UEs for the optimization of transport services, i.e.,
UEs leveraging satellite scheduled events to optimize the
communication during satellite handover. At UEs, these add-on
information helps the TCP control algorithm distinguish between
isolated packet loss and loss-inducing levels of network congestion.
Yue, et al. Expires 7 January 2026 [Page 5]
Internet-Draft The Use Cases for LEO and MEO July 2025
3.3. LEO Satellite Networks vs. Cellular Network
In this use case [Hu2023] , a measurement study was conducted to
understand the performance characteristics of Starlink satellite
networks and compare them with cellular networks. The study involved
a large-scale data collection campaign across five states in the US,
covering diverse geographical areas and user populations. The
dataset includes experimental results from both Starlink and cellular
networks, involving two types of Starlink configurations (Roam and
Mobility) and three major cellular carriers (AT&T, T-Mobile, and
Verizon). The analysis reveals several key findings:
TCP vs. UDP downlink performance. While cellular networks show
minimal TCP/UDP throughput differences, UDP dominates in satellite
environments with mean rates of 128 Mbps vs. TCP’s 29 Mbps. This
disparity stems from TCP’s reliability mechanisms (e.g., congestion
control), which amplify performance degradation under Starlink’s
elevated packet loss rates (0.3%-1.3% vs. cellular’s 0.0%-0.2%).
High loss rates trigger frequent TCP retransmissions and window size
reductions, ultimately decreasing throughput.
Uplink vs. downlink throughput. Starlink’s downlink throughput
exceeds uplink by 10x, aligning with typical user behavior favoring
data consumption (e.g., video streaming) over generation.
Latency. Starlink’s Round-Trip Time (RTT) matches cellular networks
at 50-100 ms, with only 1.8 ms (550km÷300000km/s) added per hop by
space-ground transmission, despite the 550 km satellite altitude.
Moving speed. Throughput remains stable across both networks during
high-speed movement. For Starlink, ground terminals’ mobility is
negligible relative to satellites’ 28,000 km/h speed. For cellular
networks, efficient handovers contribute to maintaining consistent
throughput.
TCP parallelism optimization. Starlink achieves a better throughput
improvement, over 50% with 4 parallel TCP connections and over 130%
improvement with 8 connections. TCP parallelism optimizes bandwidth
utilization by distributing data across multiple connections, thereby
mitigating the impact of TCP congestion control. It also improves
packet loss handling. In case of packet loss in one connection,
other connections continue data transmission, minimizing the impact
on overall throughput.
Yue, et al. Expires 7 January 2026 [Page 6]
Internet-Draft The Use Cases for LEO and MEO July 2025
In summary, Starlink experiences higher packet loss compared to
cellular networks, while latency remains similar. This high packet
loss significantly impacts TCP performance, resulting in TCP
throughput being only 1/5 of UDP throughput over Starlink. TCP
parallelism offers more benefits to Starlink than cellular networks,
likely due to its effective handling of packet loss.
To address these issues, multipath UDP experiments are conducted.
However, multipath UDP initially showed only marginal throughput
gains compared to single-path transfer. By tuning the buffer
settings, they achieved more significant throughput improvements.
The results highlight the need for better solutions that can handle
the specific characteristics of LEO networks.
4. Security Considerations
TBD.
5. IANA Considerations
This document has no requests for IANA.
6. References
6.1. Informative References
[APNIC-blog]
Huston, G., "A transport’s view of Starlink", APNIC
Blog https://blog.apnic.net/, May 2024,
<https://blog.apnic.net/（URL待补充）>.
[Hu2023] Hu, B., Zhang, X., Zhang, Q., Varyani, N., Mao, Z. M.,
Qian, F., and Z.-L. Zhang, "LEO Satellite vs. Cellular
Networks: Exploring the Potential for Synergistic
Integration", CoNEXT Companion 19th International
Conference on Emerging Networking EXperiments and
Technologies, December 2023,
<https://doi.org/10.1145/3624354.3630588>.
[I-D.yang-tsvwg-leo-transport-req]
Yang, F. and T. Tsou, "Transport Layer Protocol
Requirement for LEO satellite", Work in Progress,
Internet-Draft, draft-yang-tsvwg-leo-transport-req-00, 16
March 2025, <https://datatracker.ietf.org/doc/html/draft-
yang-tsvwg-leo-transport-req-00>.
Yue, et al. Expires 7 January 2026 [Page 7]
Internet-Draft The Use Cases for LEO and MEO July 2025
[Izhikevich2024]
Izhikevich, L., Enghardt, R., Huang, T.-Y., and R.
Teixeira, "A Global Perspective on the Past, Present, and
Future of Video Streaming over Starlink", ACM
Journal Volume 8, Issue 3, Article 30, December 2024,
<https://doi.org/10.1145/3700412>.
[RFC2119] Bradner, S., "Key words for use in RFCs to Indicate
Requirement Levels", BCP 14, RFC 2119,
DOI 10.17487/RFC2119, March 1997,
<https://www.rfc-editor.org/info/rfc2119>.
[RFC8174] Leiba, B., "Ambiguity of Uppercase vs Lowercase in RFC
2119 Key Words", BCP 14, RFC 8174, DOI 10.17487/RFC8174,
May 2017, <https://www.rfc-editor.org/info/rfc8174>.
[_GPP.SP-250400]
3GPP, "New SID on Study on Integration of satellite
components in the 5G architecture Phase 4", 3GPP SP SP-
250400, March 2025.
6.2. Normative References
[RFC2883] Podolsky, M., Floyd, S., Mahdavi, J., and M. Mathis, "An
Extension to the Selective Acknowledgement (SACK) Option
for TCP", Request for Comments 2883, DOI 10.17487/RFC2883,
publisher RFC Editor, July 2000,
<https://www.rfc-editor.org/info/rfc2883>.
[TS.23.501]
3GPP, "System Architecture for the 5G System (5GS)", 3GPP
TS 23.501 (V19.3.0), March 2025,
<https://www.3gpp.org/ftp/Specs/
archive/23_series/23.501/>.
Authors' Addresses
Shengnan Yue
China Mobile
Email: yueshengnan@chinamobile.com
Liz Izhikevich
University of California
Email: lizhikev@g.ucla.edu
Yue, et al. Expires 7 January 2026 [Page 8]
Internet-Draft The Use Cases for LEO and MEO July 2025
Tina Tsou
Tiktok
Email: tina.tsou@tiktok.com
Yue, et al. Expires 7 January 2026 [Page 9]