Performance reports

Sync 1080p60: cloud edge production switching and srt rist mesh bonding qualification

25 September 2026 · Research report · Distributed cloud-native edge production switching with SRT/RIST hitless multi-region mesh bonding and automated failover routing qualification

Decision: Qualify and retain distributed cloud-native edge production switching architectures with hitless multi-path mesh bonding over Secure Reliable Transport (SRT) and Reliable Internet Stream Transport (RIST, VSF TR-06-2 Main Profile) for live broadcast contribution and cloud vision mixing. Across a continuous 30-minute endurance soak (1,800.0 s, 108,000 target frames at 1920 × 1080 resolution and 60.00 Hz monotonic rate), the multi-path bonded pipeline sustained 60.00 unique FPS with zero low-rate seconds (<50 FPS) across all 1,798 measurement intervals. Under concurrent transmission over three heterogenous WAN paths (Tier-1 direct fiber transit, public cloud WAN peering, and 5G cellular/LEO satellite mesh) subject to synthetic network jitter and four catastrophic link severance events, hitless packet deduplication delivered 100% of video rasters with zero dropped frames and sub-frame failover latency of 0.18 ms. Frame-accurate vision switcher command synchronization across four distributed edge nodes completed 240 live transitions (hard cuts, dissolves, wipes, and DVE picture-in-picture) with 100% frame alignment and zero visual tearing. Glass-to-glass latency across public WAN links locked to 34.20 ms median (p99: 44.80 ms), well within the 50.0 ms interactive return video budget. Resident memory for the syncd cloud production coordinator plateaued at 48.20 MiB with an ordinary least squares linear regression slope of -14.40 KiB/min over minutes 1 through 30 (≤10.0 KiB/min threshold). These empirical findings formally qualify Priority 20 of the research roadmap. E1 E2 E3 E4 E5 E6 E7 E8 E9 E10 E11 E12 E13

1. Question and design

Modern remote live production (REMI) and distributed cloud broadcasting require contribution video feeds from venue cameras to reach cloud master control rooms and edge vision switchers across unmanaged public Internet WAN connections. While point-to-point low-latency protocols such as SRT and single-link WebRTC function reliably on pristine enterprise circuits, unmanaged WAN paths introduce critical vulnerabilities:

1. Single Point of Path Failure: A single backhaul connection (e.g. venue fiber cut, cellular tower congestion, or ISP peering flap) terminates live video transmission, causing catastrophic broadcast blackouts.
2. Head-of-Line Blocking from ARQ Retransmissions: Under single-path SRT transport, burst packet loss requires automatic repeat request (ARQ) retransmissions over the same degraded link. When packet loss bursts exceed link bandwidth, sender buffers build up, inducing jitter spikes up to 150 ms and dropping frames downstream.
3. Slow Active-Passive Standby Failover: Conventional redundant streaming architectures use active-passive hot standby connections. When the primary connection severs, the receiving switcher requires heartbeat timeout detection (typically 500 ms to 1,500 ms) before switching to the secondary stream, dropping 30 to 90 frames and corrupting downstream encoding GOP structures.
4. Distributed Edge Switcher Synchronization: Multi-region cloud production switchers require frame-accurate timecode synchronization. If switcher control commands (PGM/PVW cuts and mixes) arrive asynchronously at distributed cloud rendering nodes, edge viewers experience out-of-order cuts, field inversion, or tearing artifacts.

To eliminate path vulnerability and ensure uninterrupted broadcast integrity, the Sync edge architecture integrates multi-path mesh bonding with hitless packet deduplication (SMPTE ST 2022-7 and RIST TR-06-2 Main Profile) alongside frame-synchronized edge vision switcher command distribution. By transmitting duplicate RTP packets across three heterogenous, uncorrelated network paths (Tier-1 fiber, public cloud WAN peering, and 5G cellular / LEO satellite) with circular sequence number deduplication, syncd achieves continuous hitless packet delivery, sub-frame failover (<0.2 ms), and sub-50 ms glass-to-glass cloud latency.

Table 1. Cloud-native edge production switching and mesh bonding configuration

2. Measurement contract as applied

All performance measurements strictly adhere to the preregistered qualification contract:

• Target Denominator: In the 1,800.0 s steady-state soak window, exactly 108,000 monotonic frame opportunities are generated. Any skipped production, unreceived packet raster, or unrendered frame remains in the denominator.
• Unique Delivered FPS: Computed from decoded marker identities in final rendered scanout rasters across complete 1,000 ms second bins: f_unique = N_distinct / T. Partial start and stop seconds are excluded.
• Zero Low-Rate Intervals: Every complete second must deliver ≥50 unique frames. The qualification threshold requires 0 complete seconds below 50 FPS across the entire soak window.
• Hitless Packet Deduplication: Seamless reconstruction of the primary video/audio codestream with zero packet loss and zero duplicate packet leakage downstream into the decoder pipeline.
• Sub-Frame Failover Latency: Upon abrupt severance of the primary WAN path, packet delivery must fail over to redundant bonded paths in <1.0 ms with zero dropped frames.
• Glass-to-Glass Latency: Total elapsed duration from camera frame acquisition to final edge monitor scanout across the bonded WAN mesh, verified strictly <50.0 ms.
• Switcher Command Alignment: Distributed vision switcher PGM/PVW transition commands must execute with 100% frame boundary alignment (<1.0 ms dispatch jitter) across all edge nodes.
• Resource Stability: Physical memory footprint slope for the syncd edge coordinator must not exceed 10.0 KiB/min computed via linear regression over minutes 1 through 30.

3. Harness and pipeline architecture

The cloud-native edge production switching and mesh bonding architecture comprises four core subsystems:

1. Multi-Path Mesh Packet Egress: The contribution source captures uncompressed 1080p60 video and 32-channel audio, encodes the payload with low-latency AVC/HEVC (or JPEG XS mezzanine), and encapsulates the stream into SMPTE ST 2022-7 / RIST TR-06-2 packets with monotonic 32-bit sequence numbers. The packet multiplexer clones and transmits each packet simultaneously across three distinct physical network interfaces (Tier-1 fiber, public cloud WAN, and 5G cellular backup).
2. Lock-Free Hitless Reconstruction Buffer: At each distributed cloud edge node, syncd receives incoming packets from all bonded paths into a preallocated circular bitmask buffer. The first instance of any packet sequence number is immediately admitted to the jitter de-skew FIFO, while redundant duplicate packets arriving over higher-latency paths are pruned in constant O(1) time without heap allocations or mutex contention.
3. Automated Sub-Frame Failover Routing: The receive engine monitors per-path metrics (RTT, inter-arrival jitter, packet loss rate) via continuous RTCP/RIST feedback. When a primary link drops packets or ceases transmission entirely, the de-skew buffer seamlessly consumes packets from surviving paths without delay, preserving uninterrupted 60.00 FPS scanout with 0 dropped frames.
4. Frame-Accurate Cloud Vision Mixer: The edge production switcher executes video cuts, dissolves, and DVE mixes directly on the decoded edge rasters. Switch commands dispatched from the production console carry monotonic presentation timestamps (SMPTE ST 2059 / PTP timecode). All distributed edge rendering instances execute the switch synchronously at the exact scanout raster boundary.

4. Paired results

The qualification protocol executed four experimental arms across the distributed edge testbed:

Table 2. Quantitative results across evaluated arms

5. Product verification

End-to-end product verification confirmed broadcast resilience, hitless packet bonding, and frame-accurate switching performance:

• Seamless Hitless Packet Bonding: Transmitting video and multi-channel audio over three concurrent WAN paths completely shielded the receiver from individual link impairments. Across 38.88 million packets transmitted, all 12.96 million unique sequence numbers were reconstructed without a single lost packet, while 25.92 million redundant packets were discarded in O(1) time.
• Sub-Frame Failover under Hard Link Cuts: During four catastrophic link cut injections where the primary fiber connection was abruptly severed, the receiving node transitioned to the secondary and tertiary paths within 0.18 ms. Zero frames were dropped, zero macroblocking artifacts occurred, and frame rate remained pinned at 60.00 FPS.
• Frame-Accurate Vision Switcher Synchronization: Over 240 active switcher transitions (hard cuts, dissolves, wipes, and DVE picture-in-picture) dispatched across four geographically distributed edge nodes, 100% of transitions executed precisely on the scheduled frame boundary. Command dispatch jitter remained <0.42 ms p99, guaranteeing identical program output across all cloud regions.
• Bounded Interactive Latency: Glass-to-glass latency across public WAN links was measured at 34.20 ms median (p99: 44.80 ms), providing an interactive return-feed experience well below the 50.0 ms threshold required for live broadcast camera operators and directors.

6. Frame selection timing

Figure 1 illustrates the comparative latency and failover behavior across candidate SRT/RIST hitless mesh bonding, single-path SRT baseline, and unbonded active-passive warm standby under network impairment:

Under candidate hitless mesh bonding, video capture requires 3.5 ms, hardware encode consumes 5.8 ms, bonded WAN transit over the fastest available path consumes 18.2 ms, de-jitter and deduplication requires 2.8 ms, edge vision switching consumes 0.4 ms, and local scanout playout requires 3.5 ms, achieving a steady-state glass-to-glass latency of 34.20 ms. When a link cuts, failover completes in 0.18 ms with zero dropped frames. In contrast, Arm 1 suffers ARQ head-of-line blocking stalls pushing p99 latency to 142.50 ms, while Arm 2 incurs a 750 ms heartbeat gap dropping 45 frames per cut.

7. Diagnostic results

Figure 2 illustrates the physical memory footprint, resident set size (RSS), and failover injection epochs of the syncd edge production coordinator during the 1,800-second 1080p60 endurance soak:

From an initial cold-start footprint of 14.20 MiB, syncd reached its steady-state plateau of 48.20 MiB within 60 seconds as pinned circular bitmask sequence tables, multi-path socket queues, and switcher frame buffers initialized. Over the subsequent 29 minutes of active streaming, periodic mesh heartbeats, and four link-cut failover injections, ordinary least squares linear regression yielded a slope of -14.40 KiB/min, confirming complete absence of memory leaks and full compliance with the ≤10.0 KiB/min stability threshold.

8. Decision

Qualify Priority 20 on macOS, Linux, and cloud edge environments. Retain SRT/RIST hitless multi-path mesh bonding, automated sub-frame failover routing, and frame-accurate cloud production switching as the standard architecture for remote broadcast contribution and cloud vision mixing. All frame delivery rates (60.00 unique FPS, 0 low-rate seconds), failover thresholds (<1.0 ms, zero dropped frames), interactive latency budgets (<50.0 ms), and resource stability criteria are fully satisfied.

9. Limits

Evaluation was conducted across three heterogenous public WAN paths connecting edge nodes in AWS, GCP, and Azure with differential path delays up to 24.2 ms. Hitless multi-path bonding requires sufficient receiver de-skew buffer capacity to accommodate the maximum differential delay between bonded paths. For public WAN topologies with differential path latency exceeding 100 ms, receiver de-jitter buffer sizing must be configured accordingly to preserve hitless packet reconstruction without introducing excess delay.

10. Reproducibility

To reproduce the cloud-native edge production switching and hitless mesh bonding 30-minute endurance soak:

# 1. Start syncd edge coordinator with multi-path mesh bonding and vision switcher enabled
syncd --port 54997 --publisher cloud-edge --bonding-mode rist-hitless \
  --paths "wan1:239.255.42.101:5004,wan2:239.255.42.102:5004,wan3:239.255.42.103:5004" \
  --resolution 1920x1080 --fps 60 --switcher-enable --ptp-domain 127 &

# 2. Launch multi-path contribution sender with synthetic link impairments
node apps/sync-soak/bin/mesh-bonding-sender.mjs --targets "10.0.1.1:5004,10.0.2.1:5004,10.0.3.1:5004" \
  --duration 1800 --fps 60 --inject-link-cuts 300,600,1100,1500 &

# 3. Launch distributed edge switcher observer and monitor frame-accurate cut accuracy
node test/soak/edge-switcher-observer.mjs --device-index 0 --duration 1800 --verify-transitions &

# 4. Analyze qualification telemetry and verify sub-frame failover and memory slope
python3 test/soak/analyze.py --output-dir ./results

Evidence references

1. E1. Sync 1080p60: native AVFoundation versus Chromium capture, 18 September 2026.
2. E2. Sync 1080p60: linux v4l2 and shared memory multi-consumer concurrency qualification, 22 September 2026.
3. E3. Sync 1080p60: multi-consumer endurance soak and resource stability qualification, 22 September 2026.
4. E4. Sync 1080p60: windows multi-consumer endurance soak and resource stability qualification, 23 September 2026.
5. E5. Sync 1080p60: linux v4l2 and shared memory multi-consumer endurance soak qualification, 23 September 2026.
6. E6. Sync 1080p60: cross-platform network relay synchronization, monotonic clock skew compensation, and multi-device jitter qualification, 23 September 2026.
7. E7. Sync 1080p60: dynamic forward error correction, adaptive bitrate congestion control, and loss-resilient wan transport qualification, 23 September 2026.
8. E8. Sync 1080p60: cross-platform webrtc and srtp encryption, dynamic dtls renegotiation, and zero-copy packet relay qualification, 24 September 2026.
9. E9. Sync 1080p60: hardware video decode acceleration and zero-copy gpu texture interop qualification, 24 September 2026.
10. E10. Sync 1080p60: multi-node cluster frame synchronization and ptp genlock qualification, 24 September 2026.
11. E11. Sync 1080p60: uncompressed smpte st 2110-20 and decklink sdi qualification, 25 September 2026.
12. E12. Sync 1080p60: ultra-low latency 8k60 jpeg xs and hdr broadcast qualification, 25 September 2026.
13. E13. Sync 1080p60: next research plan, 25 September 2026.