May 22, 2024 The rapid evolution of Starlink’s satellite internet presents significant challenges for traditional Transmission Control Protocol (TCP), according to Geoff Huston, chief scientist at the Asia Pacific Network Information Center. In a detailed blog post, Huston explains that Starlink’s low-Earth-orbit satellites necessitate frequent hops—possibly every 15 seconds—as they rapidly traverse the sky. This constant switching introduces substantial latency variations and packet loss, which dramatically affect TCP performance.
During satellite handovers, Huston observed latency spikes and an average jitter of 6.7ms, which can severely degrade the reliability of internet connections that rely on older TCP versions like Reno TCP. These conditions create what Huston describes as “an unusually hostile link environment” for TCP, suggesting a need for protocol adaptation to maintain service quality.
To address these issues, Huston proposes tuning TCP to better suit Starlink’s unique conditions. He identifies potential solutions such as Google’s BBR (Bottleneck Bandwidth and Round-trip propagation time) protocol, which anticipates network path delays, and the CUBIC TCP congestion control algorithm paired with Selective Acknowledgement. Additionally, Explicit Congestion Notification could mitigate latency spikes during satellite transitions.
Despite these challenges, Huston notes that Starlink still delivers considerable bandwidth, citing data that shows speeds ranging from 10Mbit/sec to 370Mbit/sec. His findings highlight the need for ongoing adjustments in internet protocols to keep pace with the innovative, yet demanding, environments created by advanced satellite internet systems like Starlink.
