TSVWG                                                            F. Yang
Internet-Draft                                              China Mobile
Intended status: Informational                                   T. Tsou
Expires: 13 December 2025                                         Tiktok
                                                            11 June 2025


                    LEO transport problem statement
          draft-yang-tsvwg-leo-transport-problem-statement-00

Abstract

   LSN, Starlink and OneWeb, can provide global Internet connectivity
   with latency comparable to terrestrial networks, but their fast
   movement and frequent handovers result in highly dynamic connectivity
   changes.  The fast movement of LSN will introduce frequent handover
   and varying link delays every few minutes.  As the path over LSN
   varies, TCP cannot tell whether changes in packet loss or RTT are due
   to path changes or network congestion, thus it might not be able to
   make proper adjustment.  This greatly impact the performance of TCP.

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 13 December 2025.

Copyright Notice

   Copyright (c) 2025 IETF Trust and the persons identified as the
   document authors.  All rights reserved.

   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



Yang & Tsou             Expires 13 December 2025                [Page 1]

Internet-Draft              LEO Transport PS                   June 2025


   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
     1.1.  Terminology . . . . . . . . . . . . . . . . . . . . . . .   3
     1.2.  Requirements Language . . . . . . . . . . . . . . . . . .   4
   2.  Problem Statement . . . . . . . . . . . . . . . . . . . . . .   4
   3.  Transport Protocol over LEO Considerations  . . . . . . . . .   5
     3.1.  Congestion Control  . . . . . . . . . . . . . . . . . . .   6
     3.2.  Multi-session TCP . . . . . . . . . . . . . . . . . . . .   6
     3.3.  Less Retransmission . . . . . . . . . . . . . . . . . . .   6
   4.  IANA Considerations . . . . . . . . . . . . . . . . . . . . .   6
   5.  Security Considerations . . . . . . . . . . . . . . . . . . .   6
   6.  References  . . . . . . . . . . . . . . . . . . . . . . . . .   6
     6.1.  Normative References  . . . . . . . . . . . . . . . . . .   6
     6.2.  Informative References  . . . . . . . . . . . . . . . . .   7
   Authors' Addresses  . . . . . . . . . . . . . . . . . . . . . . .   8

1.  Introduction

   LSN, Starlink and OneWeb, can provide global Internet connectivity
   with latency comparable to terrestrial networks, but their fast
   movement and frequent handovers result in highly dynamic connectivity
   changes.  The fast movement of LSN will introduce frequent handover
   and varying link delays every few minutes.  As the path over LSN, TCP
   cannot tell whether changes in packet loss or RTT are due to path
   changes or network congestion, thus it might not be able to make
   proper adjustment.  This greatly impact the performance of TCP.

   [Izhikevich2024] by Netflix analyzed on-demand video streaming over
   LEO with one million LEO households across 85 countries for over two
   years.  Starlink users experience overall perceptual video quality
   similar to non-Starlink networks, they are more likely to experience
   bitrate switches and network rebuffers due to Starlink's network
   conditions.  About half of sessions with a bitrate switch experience
   a resolution of less than 1080p.  Video rebuffers are significantly
   more likely to occur over Starlink than a Top 10 ISP, and 40% more
   likely to occur relative to any non-Starlink network.  It also finds
   that over 95% of Starlink's throughput is lower than alternative
   networks, which is nearly always 50% of what a top 10 ISP offers and
   falls below 20 Mb/s.  This eventually leads to bitrate switching and
   rebuffer, which contributes negatively to overall quality of
   experience over Starlink.





Yang & Tsou             Expires 13 December 2025                [Page 2]

Internet-Draft              LEO Transport PS                   June 2025


   [Hu2023] investigated the performance of TCP and UDP in the Starlink
   network.  TCP vs. UDP downlink test results consistently reveal that
   UDP outperforms TCP in satellite networks, with the mean throughput
   being 128 Mbps and 29 Mbps, respectively.  And the root cause is
   there is a much higher occurrence of packet loss in both the uplink
   and downlink directions in Starlink network.  Based on this finding,
   they verified that TCP parallelism can optimize bandwidth utilization
   by distributing data across multiple connections, thereby mitigating
   the impact of TCP congestion control.

   [Li2024] reveals that the endless and bursty packet losses over
   unstable LEO satellite links impose significant challenges on
   guaranteeing the quality of experience (QoE) of Web applications.
   They found that in the Starlink network environment, both page load
   time and speed index are much higher than those in the wired network.
   Specifically, the 60th/80th/90th percentile page load time in
   Starlink are about 245.0%/202.4%/136.3% higher than those in Optimum.
   Similarly, the 60th/80th/90th percentile speed index in Starlink is
   about 185.3%/241.6%/219.3% higher.

1.1.  Terminology

   BBR: Bottleneck Bandwidth and Round-trip propagation time is a loss-
   tolerant congestion control algorithm designed by Google.

   CCA: Congest control algorithms.

   GEO: Geosynchronous orbit with the altitude 35786 km.

   LEO: Low Earth Orbit with the altitude from 200 km to 2000 km.

   MEO: Medium Earth Orbit with the altitude from 2000 km to 35786 km.

   ISL: Inter Satellite Link.

   GS: Ground Station, which connect to the satellite and provide L2
   and/or L3 functionality.

   LSN: LEO satellite networks.

   RTT: Round Trip Time.

   NTN: Non-Terrestrial Network.








Yang & Tsou             Expires 13 December 2025                [Page 3]

Internet-Draft              LEO Transport PS                   June 2025


1.2.  Requirements Language

   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.

2.  Problem Statement

   LSN differ significantly from terrestrial networks, and some of the
   design assumptions of TCP no longer hold in satellite environments.
   The unique characteristics of LSN networks-highly dynamic topologies,
   long and variable propagation delays, time-varying channel errors
   pose fundamental challenges to conventional transport layer protocols
   like TCP.  These protocols, designed for terrestrial networks, fail
   to adapt to the space environment, leading to degraded throughput,
   increased latency, and unreliable connectivity.

   *  Bursty Packet Losses: bursty packet losses are a common phenomenon
      in LEO satellite networks due to various factors such as signal
      blockage from obstacles, rapid movement of satellites, and
      interference from other signals.  These bursty losses can severely
      disrupt data transmission, leading to retransmission delays and
      increased latency.  Traditional error correction and recovery
      mechanisms may not be effective in handling such bursty losses,
      requiring the development of more robust and adaptive techniques.

   *  Under-utilized link: TCP's throughput is only about half the
      maximum possible over LSN, which is noticeably lower than that of
      the terrestrial network.  The maximum achievable throughput is
      measured through UDP bursts.  TCP throughput test has been done
      with all five congestion control algorithms: BBR, Cubic, Reno,
      Veno and Vegas.  BBR indeed achieves much higher throughput than
      the other algorithms.

   *  Variable round trip times: the dynamic nature of LEO satellite
      networks results in highly variable Round Trip Times (RTTs).  As
      satellites move rapidly across the sky, the distance between the
      sender and receiver changes constantly, causing fluctuations in
      signal propagation delay.  Additionally, network load and routing
      changes can further exacerbate RTT variations.  These
      unpredictable RTTs can lead to inefficient resource utilization
      and performance degradation in transmission protocols that rely on
      accurate timing and synchronization.






Yang & Tsou             Expires 13 December 2025                [Page 4]

Internet-Draft              LEO Transport PS                   June 2025


   *  Variable link rates: satellite-to-ground link rate is highly
      impacted by atmospheric absorption, rain attenuation, scattering
      and diffraction, ionospheric scintillation and multipath effects.
      The above factors will lead to fluctuations in the signal-to-noise
      ratio, thereby affecting the modulation mode of the channel, which
      will ultimately be reflected in the changes in the rate of the
      satellite-to-ground link.  This will create large and variable
      delay-bandwidth products.  This makes it challenging to maintain
      efficient data flow and congestion control.

   *  High bit error rates: the wireless communication environment of
      LEO satellites is prone to high bit error rates (BER) due to
      factors like signal attenuation, interference, and noise.  These
      errors can corrupt data packets, leading to retransmissions and
      reduced overall network efficiency.  Effective error detection and
      correction mechanisms are essential to mitigate the impact of high
      BER in LEO satellite transmission protocols.

   *  Out-of-order delivery: In the NTN scenario, when a mobile phone
      directly connects to a LEO (Low Earth Orbit) satellite, user
      packets are forwarded via the LSN (Link Switching Node).  Due to
      frequent link switching and faults such as sun outages, routing
      adjustments become inevitable.  This causes a single data stream
      to be forwarded over different paths, leading to packet
      reordering.

   *  TCP Fairness: there will be competing connections with different
      RTTs.  The connections with high RTTs cannot allocate enough
      bandwidth.  Traditional TCP protocols may not perform well in such
      environments due to their inherent assumptions about network
      conditions.  Developing fair and efficient resource allocation
      mechanisms that consider the unique characteristics of LEO
      satellite networks is necessary to prevent certain users or
      applications from dominating network resources.

3.  Transport Protocol over LEO Considerations

   Various optimizations at the transport protocol level have been
   proposed, such as end-to-end optimizations like SCPS-TP and M-TCP,
   redundancy coding like FEC, cross-layer optimizations like ECN, and
   congestion control algorithm optimizations like TCP Westwood and TCP
   Eifel, as well as AI-enhanced congestion control.  This gives some
   clues on what problem we should focused on.








Yang & Tsou             Expires 13 December 2025                [Page 5]

Internet-Draft              LEO Transport PS                   June 2025


3.1.  Congestion Control

   The rapidly varying satellite link capacity and latency introduced by
   LSN is unique.  Existing CCAs are unable to discriminate whether RTT/
   bandwidth variation or packet loss are caused by a congestion event
   or not.  Based on how congest events are awared, CCA can be
   classified into 4 categories, loss-based, delay-based, model-based or
   learning-based [I-D.LSNCC].  Since the packet loss has obvious burst
   loss over some period characteristics, it is possible to learn that
   by the CCA in order to isolate the congestion event more precisely.
   Especially in NTN case, the mobile terminal will be directly
   connected to satellite.

3.2.  Multi-session TCP

   Since the link is under-utilized with single TCP connection, multiple
   concurrent connections can be able to utilize a larger fraction of
   the available capacity on the path.  This has been verified by
   [Izhikevich2024] can improves overall video quality of experience.
   However, retransmit rates increase by 300% on average.

3.3.  Less Retransmission

   Since packet loss at the LSN is of the burst type, in the case of Go-
   Back-N, retransmissions will consume a large amount of bandwidth and
   increase the overall transmission completion time of the flow.  In
   the NTN scenario, when the satellite-to-ground link switches, the
   user terminal should stop sending data until the switch is completed
   to reduce retransmissions.  In non-NTN scenarios, without the
   collaboration of the LSN network, the situation is more complex.  One
   method is to automatically predict future switching time points by
   learning the regularity of burst packet loss times, thereby reducing
   the sending rate near the switching time points to minimize
   retransmissions.

4.  IANA Considerations

   N/A.

5.  Security Considerations

   N/A.

6.  References

6.1.  Normative References





Yang & Tsou             Expires 13 December 2025                [Page 6]

Internet-Draft              LEO Transport PS                   June 2025


   [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/rfc/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/rfc/rfc8174>.

6.2.  Informative References

   [Terziev2020]
              Terziev, K. and D. Karastoyanov, "The Impact of Innovation
              in the Satellite Industry on the Telecommunications
              Services Market", 2020,
              <https://api.semanticscholar.org/CorpusID:229471602>.

   [Barbosa2023]
              George, B., Sirapop, T., Beichuan, Z., and Z. Lixia, "A
              Comparative Evaluation of TCP Congestion Control Schemes
              over Low-Earth-Orbit (LEO) Satellite Networks", 2023,
              <https://doi.org/10.1145/3630590.3630603>.

   [Izhikevich2024]
              Liz, I., Reese, E., Te-Yuan, H., and T. Renata, "A Global
              Perspective on the Past, Present, and Future of Video
              Streaming over Starlink", 2024,
              <https://doi.org/10.1145/3700412>.

   [Hu2023]   Bin, H., Xumiao, Z., Qixin, Z., Nitin, V., Morley, M. Z.,
              Feng, Q., and Z. Zhi-Li, "LEO Satellite vs. Cellular
              Networks: Exploring the Potential for Synergistic
              Integration", 2023,
              <https://doi.org/10.1145/3624354.3630588>.

   [Li2024]   Jihao, L., Hewu, L., Zeqi, L., Qian, W., Yijie, L., Qi,
              Z., Yuanjie, L., and L. Jun, "SatGuard: Concealing Endless
              and Bursty Packet Losses in LEO Satellite Networks for
              Delay-Sensitive Web Applications", 2024,
              <https://doi.org/10.1145/3589334.3645639>.

   [I-D.LSNCC]
              Lai, Z., Li, Z., Wu, Q., Li, H., and Q. Zhang, "Analysis
              for the Adverse Effects of LEO Mobility on Internet
              Congestion Control", 2024,
              <https://datatracker.ietf.org/doc/draft-lai-ccwg-
              lsncc/00/>.




Yang & Tsou             Expires 13 December 2025                [Page 7]

Internet-Draft              LEO Transport PS                   June 2025


Authors' Addresses

   Feng Yang
   China Mobile
   Beijing
   China
   Email: yangfeng@chinamobile.com


   Tina Tsou
   Tiktok
   San Jose,
   United States of America
   Email: tina.tsou@tiktok.com





































Yang & Tsou             Expires 13 December 2025                [Page 8]