Low-latency applications drive an increasing number of modern applications. Latency depends on factors such as link layer technologies and how higher-layer protocols cope with transmission errors and packet losses. Most transport protocols rely exclusively on retransmissions to cope with losses with minimal overhead but potentially large tail latency. This paper leverages network coding to propose the high-speed robust tunnel (HIRT), providing timely packet delivery to any application independently of the transport protocol. A network code recovers lost data without requiring time-consuming retransmissions by adding redundancy packets, thereby slightly increasing the bandwidth usage to reduce the tail latency. Our algorithm dynamically adapts the rate of redundancy packets by measuring the network loss patterns. We implement HIRT using IPv6 Segment Routing (SRv6). We suggest an efficient software implementation and demonstrate on CloudLab that our solution can protect traffic at high speeds ($>50 \text{Gbps}$) on standard servers even when facing severe packet losses in the network. We evaluate HIRT with HTTP over TCP/QUIC, and file system benchmarks over a real network with losses, Starlink. HIRT reduces the tail latency of short HTTP requests by$2 \times$and the mean request completion time of longer requests by up to$20 \%$. HIRT also decreases the tail latency of NFS requests by up to$20 \%$.

A growing number of Internet applications require low latency. Unfortunately, most of these applications cannot use the rich features of the QUIC protocol since it only uses retransmissions to cope with packet losses. We propose, implement and evaluate QUIRL, a revisit of the QUIC loss recovery mechanism. QUIRL relies on Forward Erasure Correction (FEC) only if it is needed by the application’s latency requirements and uses classical retransmissions otherwise. We implement QUIRL and evaluate its performance for real-time video and HTTP/3. Compared to previous works adding FEC to QUIC, QUIRL is the first to be evaluated with and obtain significant performance improvements for popular applications over real lossy networks. Our evaluation shows that for video QUIRL improves the video quality while meeting strict delay requirements. For HTTP/3 transfers, QUIRL efficiently reduces the tail latency when packet losses occur without causing harm when there are no losses. We confirm these results using emulation over a wide ranges of bandwidth, delays and loss scenarios. We release our QUIRL implementation to encourage other researchers and industry to explore in more details the use of FEC in QUIC.

Packet losses are common events in today’s networks. They usually result in longer delivery times for application data since retransmissions are the de facto technique to recover from such losses. Retransmissions is a good strategy for many applications but it may lead to poor performance with latency-sensitive applications compared to network coding. Although different types of network coding techniques have been proposed to reduce the impact of losses by transmitting redundant information, they are not widely used. Some niche applications include their own variant of Forward Erasure Correction (FEC) techniques, but there is no generic protocol that enables many applications to easily use them. We close this gap by designing, implementing and evaluating a new Flexible Erasure Correction (FlEC) framework inside the newly standardized QUIC protocol. With FlEC, an application can easily select the reliability mechanism that meets its requirements, from pure retransmissions to various forms of FEC. We consider three different use cases:$(i)$bulk data transfer,$(ii)$file transfers with restricted buffers and$(iii)$delay-constrained messages. We demonstrate that modern transport protocols such as QUIC may benefit from application knowledge by leveraging this knowledge in FlEC to provide better loss recovery and stream scheduling. Our evaluation over a wide range of scenarios shows that the FlEC framework outperforms the standard QUIC reliability mechanisms from a latency viewpoint.

With new Low Earth Orbit satellite constellations such as Starlink, satellite-based Internet access is becoming an alternative to traditional fixed and wireless technologies with comparable throughputs and latencies. In this paper, we investigate the user-perceived performance of Starlink. Our measurements show that latency remains low and does not vary significantly under idle or lightly loaded links. Compared to another commercial Internet access using a geostationary satellite, Starlink achieves higher TCP throughput and provides faster web browsing. To avoid interference from performance enhancing proxies commonly used in satellite networks, we also use QUIC to assess performance under load and packet loss. Our results indicate that delay and packet loss increase slightly under load for both upload and download.

Originally implemented by Google, QUIC gathers a growing interest by providing, on top of UDP, the same service as the classical TCP/TLS/HTTP/2 stack. The IETF will finalise the QUIC specification in 2019. A key feature of QUIC is that almost all its packets, including most of its headers, are fully encrypted. This prevents eavesdropping and interferences caused by middleboxes. Thanks to this feature and its clean design, QUIC is easier to extend than TCP. In this paper, we revisit the reliable transmission mechanisms that are included in QUIC. More specifically, we design, implement and evaluate Forward Erasure Correction (FEC) extensions to QUIC. These extensions are mainly intended for high-delays and lossy communications such as In-Flight Communications. Our design includes a generic FEC frame and our implementation supports the XOR, Reed-Solomon and Convolutional RLC error-correcting codes. We also conservatively avoid hindering the loss-based congestion signal by distinguishing the packets that have been received from the packets that have been recovered by the FEC. We evaluate its performance by applying an experimental design covering a wide range of delay and packet loss conditions with reproducible experiments. These confirm that our modular design allows the protocol to adapt to the network conditions. For long data transfers or when the loss rate and delay are small, the FEC overhead negatively impacts the download completion time. However, with high packet loss rates and long delays or smaller files, FEC allows drastically reducing the download completion time by avoiding costly retransmission timeouts. These results show that there is a need to use FEC adaptively to the network conditions.

Application requirements evolve over time and the underlying protocols need to adapt. Most transport protocols evolve by negotiating protocol extensions during the handshake. Experience with TCP shows that this leads to delays of several years or more to widely deploy standardized extensions. In this paper, we revisit the extensibility paradigm of transport protocols.