Published: February 2025 | Updated March 2026
Bonding two internet connections sounds straightforward — combine a 200 Mbps fibre link with a 20 Mbps LTE line and get 220 Mbps, right? In practice, the weaker link becomes a bottleneck that can actively degrade performance rather than improve it. SD-WAN link bonding works — but only when it’s done correctly. This post explains the mechanics of packet-level bonding, why mismatched links cause problems, and what the best practices are for getting genuine bandwidth gains from a multi-link deployment.
In the world of SD-WAN, businesses are moving beyond single-link dependencies and embracing bonding technology to improve performance, resilience, and reliability.
Bonding multiple internet links together allows organisations to combine bandwidth from different sources, creating a more stable and performant network connection.
However, bonding isn’t just about slapping multiple links together and expecting magic to happen. Not all link pairings are equal, and if one of your links is significantly weaker than the other, your aggregated bandwidth will not be as efficient as you might hope.
Let’s break it down!
How Does SD-WAN Bonding Work?
SD-WAN bonding (also called link aggregation) takes multiple network links—be it fibre , LTE , wireless , or satellite —and combines them into a single logical connection. This allows traffic to be distributed across all available links in a way that maximises bandwidth and reduces packet loss.
Types of SD-WAN Bonding:
1. Packet-Level Bonding
- Splits packets across multiple links, ensuring full utilisation of available bandwidth.
- Requires careful packet reordering and jitter compensation.
2. Flow-Based Load Balancing
- Assigns different traffic flows (e.g., VoIP vs. file transfers) to different links based on performance.
- Less efficient than true bonding but simpler to implement.
3. Application-Aware Aggregation
- Uses Deep Packet Inspection (DPI) to intelligently route specific applications over the best-performing links.
- Can dynamically adjust based on link health and congestion levels.
The Limitations of Unequal Link Bonding
The dream scenario is to double or triple your bandwidth by bonding multiple links together. The reality? If one link is significantly slower than the others, it becomes a bottleneck that limits the overall benefit.
Why is this the case?
The 30% Rule: If one link is less than 30% of the bandwidth of the other, it struggles to contribute effectively in a bonded scenario.
- Example: If you bond a 200 Mbps fibre link with a 20 Mbps LTE link, the LTE connection barely adds value because packet distribution cannot be even.
- High-speed packets on the fibre link will arrive much faster than those on LTE, causing jitter and packet reordering issues.
Packet Fragmentation & Latency Issues
- SD-WAN bonding algorithms attempt to split packets evenly across all available links, but when one link is too slow, packets routed through it arrive late.
- This forces the SD-WAN appliance to reassemble out-of-sync packets, introducing delays.
- VoIP calls and real-time applications suffer the most!
Limited Aggregation Gains
- A 200 Mbps + 20 Mbps link pair does NOT equal 220 Mbps of usable bandwidth.
- Due to overheads and imbalance, you might only see 205 Mbps in real-world conditions.
What Are the Best Practices for Effective SD-WAN Bonding
To get the most out of link bonding, businesses should follow these best practices:
1. Use Links with Similar Capacities
Ideally, bonded links should be within 50-70% of each other’s capacity.
Example: A 100 Mbps link + a 70 Mbps link will bond efficiently.
But a 100 Mbps + 10 Mbps link? Not a good idea!
2. Ensure Links Have Similar Latency & Jitter Characteristics
Bonding a fibre link (5ms latency) with satellite (600ms latency) is a disaster waiting to happen.
Try to match links with similar response times to avoid packet reordering and buffer issues.
3. Configure Adaptive Bandwidth
Adaptive bandwidth in SD-WAN dynamically adjusts the amount of data sent over each link based on real-time network conditions such as latency, jitter, packet loss, and congestion.
Instead of rigidly splitting traffic, it intelligently allocates bandwidth to the best-performing links at any given moment, ensuring optimal speed, reliability, and quality for applications.
This prevents weaker links from becoming bottlenecks and improves overall network efficiency, especially in multi-link environments.
Use SD-WAN Solutions That Handle Asymmetric Bonding Smartly
✅ Not all SD-WANs are created equal!
✅ Some struggle with managing disparate links effectively.
✅ Fusion’s SD-WAN, for example, uses intelligent bonding that adapts to link conditions in real time, ensuring optimal performance.
Wrap | Balance is Key in SD-WAN Bonding
Bonding multiple links in SD-WAN is a powerful way to increase bandwidth, improve reliability, and enhance business continuity. However, not all links are created equal, and adding a significantly weaker link to a high-speed connection can negatively impact performance rather than enhance it.
By following best practices—matching link capacities, ensuring similar latencies, and using adaptive bonding techniques—businesses can fully unlock the benefits of SD-WAN without running into frustrating bottlenecks.
And remember, if your network team just throws links together without considering these factors, you might find your “bonded” setup performing worse than a single well-managed link. Don’t let that happen!
✅ Choose the right SD-WAN solution.
✅ Configure it properly.
Key Takeaways:
- True packet-level bonding splits individual packets across multiple links simultaneously — this is fundamentally different from flow-based load balancing, which assigns entire sessions to one link at a time
- The 30% rule: if one link is less than 30% of the other’s capacity, it contributes more jitter than bandwidth — the weaker link becomes a bottleneck, not a booster
- A 200 Mbps fibre + 20 Mbps LTE pair does not yield 220 Mbps — real-world usable throughput from that pairing may be only marginally above the fibre link alone
- Effective bonding requires links within 50–70% of each other’s capacity with similar latency characteristics — bonding fibre (5ms) with satellite (600ms) is counterproductive
- Adaptive bandwidth management dynamically adjusts how much traffic is sent over each link based on real-time health metrics, preventing weaker links from becoming bottlenecks
- Application-aware aggregation uses DPI to route specific applications over the best-matched link — VoIP over low-latency paths, bulk transfers over higher-capacity links
Written by
Ronald Bartels
Director: South Africa · Nepean Networks · Johannesburg, South Africa
Ronald has over 30 years of hands-on networking experience spanning financial services, ISPs, and enterprise technology. He led infrastructure at Investec for nearly eight years, managed core IP networks at iBurst, and served as a solutions architect designing data centre migrations for governments and financial institutions. Since joining Nepean Networks in 2019, he has been the driving force behind SD-WAN adoption in South Africa — engineering resilient connectivity solutions purpose-built for the realities of the local market, including load shedding, mixed-quality last mile, and infrastructure variability. Ronald holds a BSc in Computer Science from Stellenbosch University and is a Certified Data Centre Professional (CDCP).