In the early days of the internet, the primary goal of transport protocols was reliability over unstable wires. Transmission Control Protocol (TCP) achieved this through a "window-based" congestion control mechanism: send a few packets, wait for an acknowledgment (ACK), and then send more. While this works well for local networks, it fails catastrophically when data must travel across the globe or into space.
Researchers have identified a persistent bottleneck known as the "Long Fat Pipe" problem. When a network has both high bandwidth (it's "fat") and high latency (it's "long"), the time it takes for an ACK to return prevents TCP from ever reaching the link's full capacity. To solve this, a research direction has emerged focused on bypassing the standard TCP stack in favor of specialized, UDP-based transport protocols designed for high-performance wide-area networks (WANs).
The Throughput Collapse of Window-Based Control
The fundamental issue with TCP in high-latency environments is its sensitivity to packet loss. In a standard TCP implementation, a single dropped packet on a transoceanic link causes the protocol to slash its transmission rate, assuming the network is congested. Because the "round-trip time" (RTT) is so high, it takes a long time for TCP to "ramp up" back to full speed.
As Robert L. Grossman and Yunhong Gu demonstrate in their work on Sector and Sphere, standard TCP often fails to saturate high-speed links by orders of magnitude. Their solution was the development of UDT (UDP-based Data Transfer), which uses UDP as a "backplane" to implement a rate-based congestion control algorithm. Unlike TCP’s window-based approach, rate-based protocols send data at a steady, calculated pace, making them much more resilient to the high latency of transcontinental or transoceanic paths.
Saturating the Pipe for Big Science
The need for this specialized transport is most acute in "Big Science," where datasets from telescopes or particle accelerators must be moved across the globe for analysis. Chung‐Uk Lee and Seung‐Lee Kim showed that for the Korea Microlensing Telescope Network, traditional methods were insufficient for transferring 600 GB of astronomical data daily. By switching to UDP-based protocols like Tsunami and ASPERA, they could achieve speeds exceeding 500 Mbps over dedicated research networks, synchronizing remote observatories in Chile and South Africa with central hubs in South Korea in real-time.
Similarly, the evolution of Globus GridFTP, as detailed by John Bresnahan, integrated UDT support to handle the "last mile" bottlenecks of scientific data grids. These protocols treat the wide-area network not as a series of fragile hops, but as a high-performance system bus, often leveraging dedicated Layer 2 optical "lightpaths" to bypass the unpredictability of standard internet routing Grossman et al., 2007.
Solving the Latency Penalty in Space and Mobile
The problem of high latency isn't limited to scientific fiber optics; it is a defining characteristic of satellite and mobile communications. In these domains, the challenge is often "Head-of-Line Blocking"—where one lost packet stalls an entire stream of data.
James Pavur and colleagues addressed this for satellite broadband with QPEP, a Performance Enhancing Proxy based on the QUIC protocol. QUIC, which runs over UDP, allows for multiplexing multiple data streams into a single connection. If one stream loses a packet, the others continue unimpeded. This approach, combined with QUIC’s 0-RTT (zero round-trip time) handshakes, can reduce web page load times by over 50% on high-latency satellite links compared to traditional VPNs Huwyler & Pavur, 2023.
This same logic applies to Mobile Augmented Reality (MAR). Wenxiao Zhang argues in the CloudAR framework that using unreliable transport like UDP for offloading heavy computation to the cloud is superior to TCP. In AR, a delayed frame is useless; it is better to drop a packet and move to the next frame than to wait for a retransmission that would cause the user's display to stutter.
When TCP Still Wins
Despite the advantages of UDP-based transport, it is not a universal replacement. Research by Jacky Cao and Pan Hui on 5G networks suggests that when network conditions are highly optimized and latency is extremely low, the overhead of modern protocols like QUIC can actually result in higher latency than traditional TCP.
The intuition here is that UDP-based transport is a tool for adversity. It is designed to overcome the physics of long distances and the instability of wireless signals. In the "clean" environment of a 5G cell or a local data center, the simplicity of the standard TCP stack remains difficult to beat. However, as we move toward a world of global scientific collaboration and satellite-integrated 5G Xing et al., 2022, the shift toward rate-based, multiplexed UDP transport appears inevitable.