/
LeoCC Internet Congestion Control
Save to my account
Sign up
LeoCC Internet Congestion Control
LeoCC Internet Congestion Control
Study
1
Question
What is the full title of the research paper presented?
Page 2
Answer
LeoCC: Making Internet Congestion Control Robust to LEO Satellite Dynamics
2
Question
Who are the primary authors affiliated with Tsinghua University in the paper?
Page 2
Answer
Zeqi Lai, Zonglun Li, Qian Wu, Hewu Li, Xin Xie, Yuanjie Li, Jun Liu, Jianping Wu
3
Question
What CCS concepts are associated with LeoCC?
Page 2
Answer
Networks → Transport protocols; Network dynamics; Network measurement
4
Question
What are the keywords listed for the LeoCC paper?
Page 2
Answer
TCP, Congestion Control, Low-Earth Orbit Satellite Network
5
Question
How does LeoCC improve throughput compared to Cubic, Copa, and BBRv3?
Page 2
Answer
Achieves 85-494% higher throughput than Cubic, Copa, BBRv3
6
Question
Why do existing CCAs exhibit self-limited performance in LEO networks?
Page 2
Answer
They overreact to LEO-dynamics-induced network variations or overestimate link capacity causing high queuing delay
7
Question
What is the core idea behind LeoCC for handling LEO dynamics?
Page 2
Answer
Exploits 'connection reconfiguration' characteristic which correlates with network variations to discriminate dynamics-induced changes
8
Question
What mega-constellations are mentioned as deploying LEO networks?
Page 2
Answer
Starlink [8], OneWeb [9], Amazon Kuiper [4]
9
Question
What is Starlink's user base as of January 2025?
Page 2
Answer
More than five million users with Internet access in nearly 100 countries spanning all seven continents
10
Question
What differentiates LEO networks from terrestrial networks in infrastructure?
Page 2
Answer
A portion of the network infrastructure like LEO satellite switches and routers moves at high velocity relative to Earth's surface
11
Question
What network effects result from LEO satellite movement?
Page 2
Answer
Rapidly varying link capacity, frequent delay variations from path fluctuations, high packet loss rate from handovers and lossy channels
12
Question
Why do existing CCAs cause self-limited performance in LEO?
Page 2
Answer
Packet loss, delay, RTT changes mislead rate control as they assume these indicate congestion but are often dynamics-induced
13
Question
What is the conference and date for SIGCOMM 25 where LeoCC is presented?
Page 2
Answer
ACM SIGCOMM 2025 Conference, September 8-11, 2025, Coimbra, Portugal
14
Question
What is the end-to-end architecture of LEO networks as depicted?
Page 3
Answer
User device to satellite terminal to LEO satellites (with ISLs) to ground station to PoP to terrestrial Internet
15
Question
What is bent-pipe routing in LEO networks?
Page 3
Answer
Transparent forwarding when user is close to available ground station without multi-hop ISLs
16
Question
How do LEO satellites differ from GEO in orbital altitude and velocity?
Page 3
Answer
LEO at ≤2000km altitude reducing propagation delay but velocity ~7.8 km/s relative to Earth causing instability
17
Question
What causes drastic network variations in Starlink as shown in Figure 2?
Page 3
Answer
Capacity fluctuates 0-100 Mbps, RTT rapid varying or fluctuating, packet loss 0.5-6% even without full utilization
18
Question
How were network variations measured in Starlink terminals?
Page 3
Answer
Traceroute [24] and GeoIP [15] for PoP location, iperf3 [53] heavy UDP for max capacity, ICMP pings for base RTT and loss
19
Question
What performance balance do ideal CCAs aim for in LEO networks?
Page 4
Answer
Maximize throughput while minimizing delay as shown in Pareto frontier Figure 3
20
Question
Which CCAs underutilize link capacity in Starlink per Figure 3?
Page 4
Answer
Cubic, Vegas, BBRv3, Copa cannot fully utilize due to conservative response
21
Question
How does Cubic fail in replayed Starlink trace per Figure 4?
Page 4
Answer
Cannot distinguish congestion vs LEO-dynamics packet loss, shrinks cwnd conservatively causing self-limited throughput
22
Question
Why does Copa overestimate queuing delay in LEO?
Page 4
Answer
StandingRTT suddenly increases from path change but minRTT underestimates actual delay causing severe throughput drop
23
Question
What assumption does BBR violate in LEO networks?
Page 4
Answer
Highest throughput and lowest RTT over past period estimate bBW and pRTT, inaccurate due to rapid changes
24
Question
Why do learning-based CCAs like VIVACE fail under LEO dynamics?
Page 4
Answer
Best mapping from conditions to rates changes rapidly, hard to converge on time leading to under/over use
25
Question
What correlation exists between LEO reconfiguration and network variations per Figure 5?
Page 5
Answer
Capacity and RTT fit step functions divided by reconfiguration intervals, mild smooth inside, jumps on entry
26
Question
What are LeoCC's three key techniques for robustness?
Page 5
Answer
(i) Efficiently detect reconfiguration on endpoint; (ii) Reconfig-aware model to characterize/estimate net conds; (iii) Precisely regulate sending rate
27
Question
What tools record >4.8K LEO traces for LeoReplayer?
Page 5
Answer
LeoReplayer accurate record-and-replay tool for reproducible LEO environments
28
Question
What is LEO connection reconfiguration?
Page 5
Answer
Mechanism satellite operators use to schedule, manage, update connections between satellites and ground entities like terminals/GSes
29
Question
How does reconfiguration correlate implicitly with network performance?
Page 5
Answer
Updates radio frame allocation changing physical layer rate [40], causes path changes with RTT fluct and loss [46]
30
Question
What indicates reconfiguration events in Starlink terminals per Figure 7?
Page 6
Answer
RI outliers > Δ_outage (~45ms) from 10ms ICMP pings to PoP, short outage 45-120ms at network layer