Performance Test Results
Independent benchmarks validating CloudWing FEC acceleration against commercial-grade satellite optimization solutions
Executive Summary
Comprehensive testing conducted on July 10-19, 2026, comparing CloudWing's adaptive FEC (Forward Error Correction) technology against commercial-grade Hybla dual-end PEP (Performance Enhancing Proxy) and cellular backhaul optimization scenarios.
Small File Transfer
2.7~4.4×
faster than Hybla PEP in rain fade conditions
Large File Throughput
1.7×
improvement in moderate rain fade (8% loss)
Packet Loss Recovery
13.6% → 0.0%
input loss reduced to zero output loss
Key Finding
While commercial Hybla PEP represents the current state-of-the-art in pure software satellite optimization, CloudWing FEC demonstrates consistent advantages across all test scenarios—particularly in user-perceived latency for web/application interactions, where the advantage ranges from 2.7× to 4.4× over Hybla.
Test Methodology
Link Simulation Parameters
| One-way Latency | 275ms (fixed) |
| Round-trip Time (RTT) | ≈600ms |
| Uplink Bandwidth | 2 Mbit/s |
| Downlink Bandwidth | 8 Mbit/s |
| Loss Model | Uniform random (Bernoulli) |
Calibrated against typical GEO satellite characteristics (ITU-R S.1453): orbital height 35,786km, signal propagation delay 250-290ms one-way.
Test Scenarios
Clear Sky (0% packet loss)
Baseline performance with no artificial degradation
Rain Fade 8 (8%/4% bidirectional loss)
Moderate rainfall, typical rain fade condition lasting tens of minutes
Rain Fade 16 (16%/8% bidirectional loss)
Heavy rainfall, critical rain fade threshold where commercial modems begin service degradation
Competing Systems
CloudWing qpep-FEC
- • Adaptive RS (Reed-Solomon) erasure coding
- • T0: No FEC redundancy (clear sky)
- • T1: 8:2 FEC ratio (recovers up to 20% loss)
- • T2: 4:2 FEC ratio (recovers up to 33% loss)
- • Automatic gear shifting based on real-time loss detection
Hybla Dual-End PEP (Benchmark)
- • TCP split architecture (local termination)
- • ACK spoofing for immediate window opening
- • Hybla congestion control on both ends
- • Represents commercial-grade satellite modem PEP capability
- • Close to current state-of-the-art in pure software PEP
Test Results: Hybla PEP Comparison
| Scenario | CloudWing FEC (median) | Hybla PEP (median) | FEC Advantage |
|---|---|---|---|
| Clear (0% loss) | 0.710s | 1.886s | 2.66× |
| Rain8 (8%/4% loss) | 0.733s | 2.445s | 3.34× |
| Rain16 (16%/8% loss) | 1.423s | 6.294s | 4.42× |
Critical Observation: Stability
In Rain16 scenario, Hybla PEP exhibited extreme variance: samples ranged from 3.3s to 12.5s (3.8× internal variance), making user experience completely unpredictable. CloudWing FEC maintained tight distribution: 1.20s to 3.24s, providing consistent, expected performance.
Why This Matters
Small file transfer time directly represents web page loading experience and application interaction latency. A typical web page consists of dozens of small objects (HTML, CSS, JavaScript, images), each requiring separate HTTP requests.
With Hybla PEP, each object risks TCP RTO (Retransmission Timeout) penalties of 600ms+ per丢包 event. With FEC, packet loss is eliminated at the link layer before TCP ever sees it.
Technical Mechanism
Hybla/BBR: Optimizes "how fast to recover after packet loss" — but recovery still costs ≥1 RTT (600ms+).
FEC: Eliminates "packet loss itself" through redundant coding. The QUIC layer sees a near-0% loss link, so no RTO is ever triggered.
Cellular Backhaul Validation
Use Case: 4G/5G Base Station Satellite Backhaul
In cellular backhaul scenarios, user traffic is encapsulated in GTP-U (UDP port 2152) and signaling uses SCTP (protocol 132). Commercial modem PEP and standard TCP acceleration (like qpep) are ineffective because they cannot see inner TCP connections. L3 FEC tunnel is the only viable optimization method.
GTP-U/SCTP Precision分流
✅ Verified: GTP-U packets enter FEC tunnel, non-backhaul traffic bypasses untouched
ROUTED Mode (NAT-free)
✅ Verified: Tunnel preserves original source IP end-to-end, critical for SCTP multi-homing
FEC vs Rain Fade
✅ Verified: 13.6% input loss → 0.0% output loss, 10,374 packets recovered
Deployment Safety: Zero Impact on Live Network
- • Bump-in-the-wire topology: Device sits transparently between base station and satellite modem
- • No configuration changes: Base station and core network require zero modifications
- • Fail-safe bypass: Single command (`backhaul_route.sh disable`) instantly removes device from path in seconds
- • Protocol-specific handling: Only GTP-U/SCTP enters tunnel; management/OAM traffic passes through unchanged
Phased Validation Roadmap for Operators
| Phase | Operator Resource | Live Network Risk | Key Deliverable |
|---|---|---|---|
| 0: Lab/Simulation | Zero | None | Technical feasibility materials |
| 1: Mirror Port Monitoring | One SPAN port | Zero (read-only) | Real traffic processing validation |
| 2: Single Site Pilot | One low-traffic base station | Low (second-level rollback) | Rain fade KPI improvement evidence |
| 3: Gray Scale → Cutover | Batch rollout | Controlled (phased + rollback) | Full-scale deployment |
Operators can start from zero-risk mirror port monitoring (Phase 1), validating with real backhaul traffic before committing to any live deployment. Each phase has clear entry criteria, success metrics, and rollback procedures.
Technical Positioning
Complementary Relationship
CloudWing FEC operates at link layer (L3), eliminating packet loss before TCP ever perceives it. Commercial PEP operates at transport layer (L4), optimizing how TCP responds to perceived loss.
This creates a pure positive叠加 relationship: CloudWing + Commercial Modem > Commercial Modem alone. In clear sky, FEC shuts off (T0 gear), allowing modem PEP to work normally. In rain fade, FEC eliminates loss at link layer, preventing TCP from entering RTO backoff cycles.
Differentiation vs Patent Solutions
• Encrypted GTP-U: Hughes/Gilat require patent-level IPsec splitting (US10,749,667); CloudWing is natively transparent
• Deployment: Commercial solutions require modem integration/replacement; CloudWing is external bump-in-the-wire
• Investment impact: Patent solutions need modem upgrade; CloudWing has zero impact on existing investment
Honest Limitations & Pending Validation
Test Scope Boundaries
- • Burst loss (Gilbert-Elliott model) not yet tested vs Hybla — theoretically FEC advantage diminishes but remains positive
- • netem cannot simulate ACM (Adaptive Coding Modulation) dynamics in real modems
- • Physical layer ARQ in commercial modems may provide additional protection not modeled here
Pending Real-Sky Validation
- • SCTP association survival under 600ms+ RTT with rain fade (HEARTBEAT persistence over 3-minute rain event)
- • End-to-end ROUTED mode with real operator IP blocks and static return routing
- • Precise FEC on/off A/B testing for GTP-U tunnel under high-pps continuous flow
Overall Conclusion
Effectiveness (Against Link Degradation)
- ✓ Rain fade resistance: FEC reduces 13.6% burst loss to 0.0% output loss
- ✓ Protocol transparency: Works on GTP-U/SCTP/IPsec without inner protocol awareness
- ✓ User experience: 2.7~4.4× faster web/app interactions in rain fade
Deployment (Operator-Friendly)
- ✓ Bump-in-the-wire: Zero configuration changes to base station or core network
- ✓ Fail-safe bypass: One-command, second-level无损 rollback
- ✓ Phased validation: Start from zero-risk mirror monitoring, progress at your pace
CloudWing FEC provides irreplaceable value in sustained rain fade and multi-small-object scenarios where Hybla/BBR optimization cannot prevent TCP RTO penalties. The adaptive design ensures "no penalty in clear sky, deterministic gain in rain fade."