Mobile Link Bonding Comparison

Study overview

These tests evaluated foundational performance differences between two platforms commonly deployed in mobile link bonding applications. The study examined the bandwidth required to maintain bonded connections, how efficiently each platform used available capacity, and what happened when one active link was interrupted.

Introduction & test environment

The test environment was designed to minimize variability between devices and test runs. Cellular connectivity was disabled on both edge devices. Each device received the same connectivity over Gigabit Ethernet. Initial tests used unrestricted 1 Gbps access; bandwidth-limited tests used Linux Traffic Control (tc) to rate-limit access.

The Dejero GateWay M6E6 and Peplink BR2 Pro 5G were selected because of their common deployment in mobility applications.

For hub reassembly, Dejero Concentrator and Peplink FusionHub software were installed on a virtualization hypervisor with 10 Gigabit access to the test network. CPU utilization was monitored throughout testing to check for processing limitations.

For the controlled overhead and throughput tests, data remained within the lab network to avoid Internet instability influencing the results. The link failure scenarios described later used a terrestrial fiber connection and a LEO satellite connection.

Only standard, documented settings were used to adjust performance and behavior. Advanced features and hidden menus were not used.

Overhead evaluation

Link bonding introduces overhead by distributing data across multiple connections and reassembling it so that ordinary network applications can use the resulting connection. This evaluation separates overhead into three categories.

Idle link overhead

Bandwidth used to maintain connectivity, even when no client data is traversing the network.

Data transfer overhead

Additional data transmitted over the underlying WAN links relative to the client data entering the solution.

Available bandwidth utilization

Achievable throughput through the solution compared with the combined available capacity of its links.

Idle link overhead

No client devices were connected during this test. The LAN environment on each edge device was physically disconnected.

Dejero SBT does not provide a user adjustment for the aggressiveness of link failure detection, so one idle overhead configuration was tested. Peplink SpeedFusion offers four Link Failure Detection Time settings. Shorter detection times require more health checks and increase bandwidth overhead.

Dejero’s 65 MB daily overhead was greater than Peplink Recommended and lower than Peplink Fast. The operational significance of this tradeoff is explored in the link failure tests.

Data transfer overhead

Data transfer overhead measures the extra data transmitted over the WAN links. For example, transmitting 1.25 GB over the WAN to transfer a 1 GB file represents 0.25 GB, or 25%, overhead.

Using iperf3, 10 GB of TCP data was uploaded, followed by 10 GB downloaded. This sequence was repeated twice on each platform. Overhead was calculated from the difference between bytes entering the LAN port and total bytes exiting the WAN ports, with the measurements averaged to produce an overhead percentage.

The supplied test summary describes this method but does not include numerical results for data transfer overhead.