Appear Unveils Hardware-Accelerated SRT

Appear says its hardware-accelerated approach to Secure Reliable Transport (SRT) removes scaling limits for live contribution and distribution, enabling hundreds of services and UHD workflows over the public internet without building out a sprawling server estate.

SRT has become a fixture in live media for moving high-quality video across unpredictable networks, offering an alternative or complement to satellite, dedicated fibre and leased circuits. As usage grows, the question has shifted from whether SRT can carry a feed to whether it can sustain hundreds of services, high-bitrate UHD contribution and mission-critical operations without excessive infrastructure.

Appear's initial deployments relied on the open-source LibSRT library, processing the protocol entirely on CPUs. The company has since reworked its implementation to offload media transport into hardware, separating SRT’s control plane on the CPU from the data plane on an FPGA. That split, Appear argues, puts each task on the component best suited to it and unlocks significantly higher density.

In practice, Appear says a single IP gateway module now supports up to 128 services and 6 Gb/s of SRT traffic, with retransmission bandwidth controlled per service. Configured as a dedicated gateway in 2RU with 100Gb input connectivity, an X20 chassis scales to 1,536 simultaneous SRT connections and 72 Gb/s of aggregate throughput. By the company’s nominal service-rate assumptions, that equates to roughly 72 UHD JPEG XS services at 1 Gb/s, or around 900 UHD HEVC services at 80 Mb/s. In a combined HEVC encoding-and-SRT setup, the platform can carry up to 22 concurrent UHD camera feeds.

According to Appear, the hardware offload reduces packet-processing overhead and delivers predictable capacity, while overall latency remains governed by buffer settings and prevailing network conditions.

The company frames the shift as an economics story as much as a performance one. While small-scale use cases may be best served in software—its VX platform uses a software implementation of Appear’s SRT stack for a handful of cloud-ingest feeds—the calculus changes as operations grow. Server estates bring costs in rack space, power, cooling, operating-system maintenance, patching and additional failure points. FPGA acceleration, Appear contends, consolidates workloads into fewer boxes with known capacity per module, allowing operators to scale by adding modules rather than servers.

Appear positions its product families accordingly. X5 targets compact AVC/HEVC processing with SRT transport at the edge, including JPEG XS over SRT for latency-sensitive applications. X10 and X20 address higher-density, modular deployments demanding deterministic performance at both the edge and the core. VX remains the software option when dedicated hardware is unnecessary.

Recent customer projects illustrate how the approach maps to real-world conditions. In Chile, Appear says hardware-accelerated SRT carried an HEVC contribution feed from Chilevisión into AWS to support cloud processing for a nationwide 4K FIFA World Cup service. World Archery has shifted to sending SRT feeds from global venues to its production centre in Lausanne and then on to more than 15 media partners; the federation calculated an 18‑month payback after bringing production in-house. In Canada, Corus Entertainment is using SRT as its primary transport protocol across affiliates, news bureaux and operational hubs, reducing reliance on leased circuits while retaining legacy interfaces where needed.

The company also seeks to counter common assumptions about SRT at scale. First, it argues that large SRT deployments do not inherently require vast server farms—one IP gateway module can host up to 128 services and 6 Gb/s, and a 2RU X20 can reach 1,536 connections and 72 Gb/s, reducing rather than multiplying server counts. Second, it maintains that software-based SRT is not the wrong choice by default; at small scale, it is often the right one, with hardware acceleration earning its place when density and deterministic performance become priorities. Finally, Appear emphasises that the discussion is fundamentally about total cost of ownership—rack space, power, cooling and patching—rather than raw throughput alone.

Looking ahead, Appear says its SRT implementation already supports bonding, rendezvous mode and ST 2022-7 seamless protection switching, with deeper monitoring a natural next step. The underlying proposition remains straightforward: reliable, secure transport across widely available networks, extended through hardware acceleration from a few remote feeds at the edge to high-density contribution and distribution at the core.

www.appear.net