SpeedFusion Configuration
SpeedFusion is what turns a Peplink router into something no other vendor can match. Multiple WAN connections, bonded into a single resilient path. We configure it properly so your organisation gets every bit of performance and reliability the technology can deliver.
What SpeedFusion Actually Does
Every Peplink router can connect to multiple WAN sources simultaneously. That part is straightforward. You might have a fibre line, two 5G SIMs and a Starlink terminal all plugged into the same device. Without SpeedFusion, the router can use those connections for load balancing and basic failover. Useful, but limited. Traffic goes out on one link at a time. If that link drops, sessions break while the router switches to another.
SpeedFusion changes this entirely. It builds an encrypted tunnel between two Peplink endpoints and bonds the underlying WAN connections into a single logical path. Traffic is split across every available link at the packet level, reassembled at the far end, and delivered to the destination as though it came from one connection. The result is aggregated bandwidth, unbreakable sessions and the ability to survive individual link failures without any user noticing.
This is not theoretical. Broadcast crews bond four or five cellular connections to push live HD video from a moving vehicle. Maritime operators combine VSAT with multiple 4G connections to maintain operational communications while at sea. Enterprise organisations bond FTTP, LTE and Starlink at branch offices to guarantee uptime for VoIP, video conferencing and cloud applications. The technology works. But it needs to be configured correctly for each use case, and that is where we come in.
As a Peplink Gold Partner, our engineers configure SpeedFusion tunnels for production environments every week. Broadcast, maritime, events, defence, enterprise. After hundreds of deployments, we understand the settings that the documentation glosses over, the parameter interactions that only surface under load, and the firmware-specific behaviours that can trip up even experienced network engineers.
Aggregate Every Connection into One
Bandwidth bonding is the headline capability of SpeedFusion. It takes every WAN connection on the device, splits outbound traffic across all of them, and reassembles it at the remote endpoint. The aggregate throughput equals the sum of all links, minus a small overhead for the tunnel encapsulation. If you have a 50 Mbps fibre line, a 30 Mbps 5G connection and a 20 Mbps Starlink link, your bonded tunnel delivers close to 100 Mbps of usable throughput.
The bonding algorithm distributes packets intelligently, accounting for the latency and available bandwidth of each link in real time. More packets go down faster links, fewer down slower ones. When a cellular connection drops from 30 Mbps to 8 Mbps because signal quality has degraded, the algorithm adjusts within seconds. No manual intervention. The tunnel keeps running, and throughput adjusts smoothly rather than dropping to zero while a failover event processes.
Configuration requires careful attention to bandwidth allocation weights, tunnel encryption settings (AES-256 by default, but sometimes reduced to AES-128 or disabled entirely for broadcast applications where throughput matters more than encryption), and WAN health check intervals. The health check is critical. Set it too aggressively and you generate unnecessary overhead. Set it too loosely and the tunnel takes too long to detect a degraded link. One-second health checks are a good starting point for cellular-heavy deployments; we tune from there based on observed behaviour.
Bonding requires an endpoint at the far side. Traffic from your edge device must terminate somewhere that can receive the SpeedFusion tunnel, unbundle the packets and forward them to the internet or your private network. That endpoint can be a physical Peplink device at your data centre, a FusionHub virtual appliance running on AWS, Azure or on-premises infrastructure, or SpeedFusion Cloud, which is Peplink's own hosted bonding endpoint service. Each option has trade-offs around cost, latency, control and throughput capacity. We help you choose the right one for each deployment.
Sub-Second Recovery When a Link Drops
Standard WAN failover on a conventional router takes anywhere from 10 to 60 seconds. During that window, VoIP calls drop, VPN sessions disconnect, video streams buffer and cloud applications time out. Users notice. Operations stop. When the failover finally completes, most applications need to be manually reconnected.
SpeedFusion hot failover works differently. Because the bonded tunnel already spans all WAN connections simultaneously, the loss of a single link does not break the tunnel. Remaining links continue carrying traffic without interruption. There is no failover event in the traditional sense; the tunnel simply adjusts its packet distribution to exclude the failed link and increases the load on the survivors. Switchover happens in under one second, often in under 200 milliseconds. VoIP calls stay connected. Video conferences continue without dropping.
This is particularly valuable in cellular environments where individual links are inherently unreliable. A 4G connection might drop momentarily as a vehicle moves between cell towers. A Starlink terminal might lose signal briefly during heavy rain. With SpeedFusion, these interruptions pass unnoticed. The tunnel absorbs them, and the user experience remains consistent.
Priority settings match the cost and performance characteristics of each link. A cheap 4G connection might be set as always-on, while an expensive VSAT link is held in reserve and only activated when the primary links are insufficient. The router monitors each WAN connection continuously and brings links in or out of the tunnel based on configurable health check thresholds for latency, packet loss and throughput.
Eliminate Jitter and Packet Loss
Bonding and failover handle the big problems: total throughput and link failure. WAN Smoothing and Forward Error Correction (FEC) handle the subtle ones: jitter and packet loss. These are the issues that make VoIP calls sound choppy, video feeds pixelate and real-time applications feel sluggish, even when the connection appears to be working.
WAN Smoothing works by sending duplicate copies of each packet across multiple WAN links simultaneously. The receiving endpoint accepts whichever copy arrives first and discards the duplicates. The effect is dramatic. If one link experiences a 100ms latency spike, the duplicate arriving on a different link at the normal 30ms latency is used instead. The application sees consistent, low-latency delivery regardless of how individual links are performing. The trade-off is bandwidth. Sending packets two or three times consumes two or three times the bandwidth for that traffic. This is why we configure WAN Smoothing selectively, applying it only to latency-sensitive traffic such as VoIP, video conferencing and real-time control data, while bulk traffic like file transfers and web browsing uses standard bonding.
Forward Error Correction takes a different approach. Rather than duplicating entire packets, FEC adds mathematical redundancy to the data stream. If 5% of packets are lost in transit, FEC can reconstruct the missing packets from the redundancy data without needing a retransmission. This is faster than waiting for TCP to detect the loss and request a resend, which adds at least one round-trip time of delay. FEC is particularly effective on satellite connections, where round-trip times exceed 500ms and retransmission delays are painful.
FEC levels are tuned based on the observed packet loss rate of each deployment. A terrestrial cellular connection in an urban area might need minimal FEC, perhaps 10% redundancy. A VSAT connection on a vessel in the North Sea might need 30% or more. Over-provisioning wastes bandwidth. Under-provisioning lets packet loss through to the application layer. Getting the balance right requires testing under real conditions, which is part of every SpeedFusion deployment we commission.
Site-to-Site VPN Tunnels Between Peplink Devices
PepVPN is the foundation that SpeedFusion builds upon. It creates encrypted site-to-site VPN tunnels between any two Peplink devices, forming the transport layer for all SpeedFusion features. Every Peplink router supports PepVPN, including the entry-level models. The key distinction: SpeedFusion features like bonding, WAN Smoothing and FEC require a SpeedFusion licence, while basic PepVPN tunnel connectivity ships with the hardware.
PepVPN tunnels support Layer 2 and Layer 3 modes. Layer 2 tunnels bridge remote LANs as though they were on the same physical network, useful for applications that rely on broadcast traffic or need devices at different sites to appear on the same subnet. Layer 3 tunnels route traffic between different subnets at each site, and this is the more common configuration for enterprise deployments. The right mode depends on the applications running across the tunnel and the network architecture at each site.
For multi-site organisations, PepVPN tunnels can be configured in hub-and-spoke or full-mesh topologies. Hub-and-spoke routes all inter-site traffic through a central hub, which simplifies management but introduces a single point of failure. Full mesh connects every site directly to every other site, eliminating the hub bottleneck but increasing the number of tunnels to manage. In practice, most organisations are best served by a hybrid: hub-and-spoke for general internet access with direct mesh tunnels between sites that exchange heavy traffic, such as shared file servers or video production workflows.
Each tunnel is encrypted using AES-256 by default. We configure tunnel priorities, bandwidth limits and routing rules to ensure that business-critical traffic receives precedence. Quality of Service (QoS) policies within the tunnel allow VoIP traffic to be prioritised over bulk data transfers, ensuring call quality remains high even when the tunnel is under load.
SpeedFusion Cloud and FusionHub
Every SpeedFusion tunnel needs two endpoints. The edge device is the Peplink router at your site. The remote endpoint is where the bonded traffic terminates and gets forwarded to the internet or your private network. You have three options for this remote endpoint, and choosing the right one is a critical design decision.
SpeedFusion Cloud
SpeedFusion Cloud is Peplink's managed bonding endpoint service. You do not need to deploy any hardware or virtual machines. Peplink operates a global network of data centres that act as your tunnel endpoints. You select the nearest location, point your device at it, and the tunnel comes up automatically. SpeedFusion Cloud is sold as a subscription, typically bundled with the device or available as an add-on licence. It is the simplest option and works well for smaller deployments, remote workers and situations where you need bonding but do not want to manage infrastructure. The limitation is that you share the endpoint with other users, and you have less control over the routing and throughput capacity.
FusionHub
FusionHub is a virtual appliance that you deploy on your own infrastructure. It runs on VMware ESXi, Hyper-V, KVM, AWS EC2, Microsoft Azure, Google Cloud and several other hypervisors. FusionHub acts as a dedicated SpeedFusion endpoint, giving you full control over the tunnel termination point, the routing configuration and the throughput capacity. For organisations with existing data centre infrastructure or cloud accounts, FusionHub is usually the better option. You can place FusionHub instances in multiple regions for geographic redundancy, and you can scale throughput by deploying larger virtual machine sizes.
We deploy and configure FusionHub instances as part of our SpeedFusion projects. This includes sizing the virtual machine correctly (CPU and memory requirements scale with the number of tunnels and the aggregate throughput), configuring the cloud provider's networking (security groups, elastic IPs, route tables), and tuning the FusionHub settings to match the edge devices connecting to it. For high-availability deployments, we configure FusionHub pairs with automatic failover so that the loss of one instance does not affect tunnel connectivity.
Physical Hub Device
The third option is a physical Peplink device at your data centre or head office acting as the hub. A Balance 380x or 580x, for example, can terminate dozens of SpeedFusion tunnels from remote sites. This is common in enterprise SD-WAN deployments where the organisation already has rack space and internet connectivity at a central location. We configure the hub device with the appropriate SpeedFusion licence tier, set up the tunnel profiles and deploy InControl2 templates so that new edge devices connect automatically when powered on.
Best Practices We Follow on Every Deployment
SpeedFusion has dozens of tuneable parameters. The difference between a tunnel that performs brilliantly and one that wastes bandwidth or drops sessions under load often comes down to a handful of settings that interact in non-obvious ways. These are the configuration areas where our experience makes the most difference.
Bandwidth Allocation and Weighting
Each WAN connection in a bonded tunnel can be assigned a bandwidth weight. If your fibre line delivers 80 Mbps and your 4G connection delivers 20 Mbps, setting the weights to 80:20 ensures the algorithm distributes packets in the correct proportion. Getting this wrong results in the tunnel buffering packets on the slower link, which increases latency for all traffic. Most people set weights once and forget them. That is a mistake. Actual throughput on each WAN connection should be measured before setting weights, and revisited after a week to account for real-world variation.
Tunnel Priorities and Traffic Routing
Not all traffic should go through the SpeedFusion tunnel. DNS lookups, NTP queries and low-priority web browsing can use standard WAN load balancing without the overhead of tunnel encapsulation. Save the bonded tunnel for what actually matters: VoIP, video conferencing, remote desktop, cloud ERP. Outbound policies and traffic rules direct each type of traffic to the appropriate path, keeping tunnel overhead low while protecting the traffic that counts.
Health Check Intervals
The WAN health check determines how quickly the tunnel detects a degraded or failed link. Never rely on a single probe method; ICMP, HTTP and DNS probes in combination eliminate false positives. Check interval is typically one second for cellular deployments, two seconds for fixed-line connections. The failure threshold (how many consecutive failed checks before a link is removed from the tunnel) matters just as much. Too low and a momentary blip causes unnecessary churn. Too high and a genuinely failed link stays in the tunnel, dragging down performance.
Encryption and Overhead
SpeedFusion tunnels are encrypted by default using AES-256. For most enterprise deployments, this is the correct setting. For broadcast applications pushing 50 Mbps or more of bonded video, the encryption overhead can consume meaningful CPU cycles on the router, reducing maximum throughput. In these cases, we evaluate whether AES-128 or no encryption is acceptable, weighing the security requirements against the throughput needs. Newer Peplink models with hardware encryption acceleration handle AES-256 at line rate, making this a non-issue on current hardware.
Tunnel Keepalive and Reconnection
When a tunnel drops entirely, such as when both the edge device and all WAN links go offline during a power failure, the reconnection behaviour matters. We configure aggressive keepalive intervals so the tunnel re-establishes within seconds of connectivity being restored. For mobile deployments where devices frequently power cycle, we configure persistent tunnel profiles in InControl2 so the device reconnects to the correct FusionHub or SpeedFusion Cloud endpoint automatically without manual intervention.
Real-World SpeedFusion Deployments
Broadcast Uplink Bonding
A broadcast production company needs to transmit live HD video from outdoor locations with no fixed connectivity. We deploy a Peplink HD4 with four cellular SIMs from different carriers, bonded through SpeedFusion to a FusionHub instance in a London data centre. WAN Smoothing is enabled on the video encoder's traffic to eliminate jitter. The bonded throughput of 60 to 80 Mbps is sufficient for a 1080p feed with headroom. When the crew moves between locations, the cellular connections hand off between towers and the bonded tunnel adjusts automatically. The gallery receives a clean, continuous feed regardless of how individual cellular links are performing.
Maritime VSAT and Cellular
A vessel operating in European coastal waters has VSAT for open-ocean connectivity and dual 4G for coastal coverage. We configure SpeedFusion to bond all three connections when available, with the 4G links prioritised for cost and latency. When the vessel moves beyond cellular range, the tunnel continues on VSAT alone. When it returns to coastal waters, the 4G links re-enter the tunnel automatically. VoIP traffic uses WAN Smoothing to maintain call quality over the high-latency VSAT link. Crew welfare internet is routed outside the tunnel to avoid consuming bonded bandwidth. The bridge systems, navigation data and operational communications run through the bonded tunnel with FEC enabled to handle the packet loss typical of maritime satellite connections.
Enterprise SD-WAN with Branch Office Bonding
An organisation with 40 branch offices across the UK needs resilient connectivity for cloud-hosted ERP, VoIP and video conferencing. Each branch has an FTTP connection and a 5G backup on a Peplink B One 5G. All branches connect via SpeedFusion to a pair of FusionHub instances running on AWS in eu-west-2 (London). The FusionHub pair provides redundancy; if one instance fails, all tunnels fail over to the second within seconds. VoIP traffic is tagged for WAN Smoothing. ERP traffic is routed through the bonded tunnel. General web browsing breaks out locally to reduce tunnel load. InControl2 manages the entire estate from a single dashboard, with firmware updates staged across groups and alerts configured per-site. The result is enterprise-grade SD-WAN built entirely on Peplink hardware, without the licensing complexity and per-seat costs of traditional SD-WAN vendors.
What You Get from a SpeedFusion Engagement
- Full SpeedFusion tunnel configuration on all edge devices and hub endpoints
- Bonding, failover, WAN Smoothing and FEC settings tuned to your specific use case
- FusionHub deployment and configuration on AWS, Azure or on-premises infrastructure
- SpeedFusion Cloud provisioning and device pairing where appropriate
- PepVPN tunnel architecture: hub-and-spoke, mesh or hybrid topology
- Traffic routing rules and outbound policies separating tunnel and non-tunnel traffic
- QoS configuration prioritising VoIP, video and business-critical applications
- WAN health check tuning with appropriate probe types and failure thresholds
- InControl2 group configuration and tunnel monitoring dashboards
- Load testing under real conditions to validate throughput and failover times
- Documentation covering tunnel architecture, settings and troubleshooting procedures
- Handover and training for your team to manage day-to-day tunnel operations
Every engagement starts with understanding your operational requirements. There is no universal SpeedFusion template. Configuration that works for a broadcast uplink truck is entirely different from configuration that works for a 40-site enterprise SD-WAN. The technology is the same; the tuning is specific to your environment, your traffic patterns and your tolerance for risk.
Services That Work Alongside SpeedFusion
Ready to Configure SpeedFusion?
Whether you need a single bonded uplink or a nationwide SD-WAN deployment, our engineers will design and configure the SpeedFusion architecture that fits your requirements. Tell us about your project and we will come back with a plan.
Discuss Your SpeedFusion Project