Multipath Reliable Connection Protocol

When two devices exchange data, they adhere to fixed rules. Such sets of rules are called protocols. They determine how data is broken down into small packets, sent, and reassembled at the receiver. A Multipath Reliable Connection Protocol is such a set of rules with two special properties. First, it uses multiple connection paths simultaneously, for example WiFi and mobile network. Second, it guarantees that all packets arrive in the end, and in the correct order. The phone, meanwhile, sees only a single, stable connection on the outside.

Why one path is rarely enough

Most devices have long had multiple network access points. A smartphone has WiFi and mobile network, a laptop additionally has a cable connection. Classic protocols, however, always use only one of these. If you switch from WiFi to mobile network when leaving your apartment, the old connection breaks off. A video call then freezes for a few seconds or ends entirely.

Multipath protocols solve this problem because they keep both access points open in parallel. If the WiFi disappears, the mobile network immediately takes over the remaining packets. The switch happens within milliseconds and remains invisible to the app. This is exactly what the word “reliable” refers to.

There is also a second advantage: more speed. Two paths can add their bandwidth together, similar to how two open checkout counters in a supermarket clear the queue faster. Large downloads or cloud backups benefit noticeably from this. This is relevant for AI applications, because voice assistants and video analyses constantly send data to data centers.

Distributing packets and sorting them again

At the start, the protocol establishes a main connection, for example via WiFi. It then reports further available paths to the counterpart. Several individual connections thus become a shared session. Every data packet receives a sequential number before it is sent off.

A distributor then decides for each packet which path it takes. This decision depends on how fast and how utilized the individual paths currently are. The protocol continuously measures response times and adjusts the distribution accordingly. A slow mobile network path therefore receives less load than a fast WiFi connection.

At the receiver, the packets arrive out of order because the paths have different speeds. There, a buffer sorts them back into the correct order using the numbers. If a packet is missing, it is requested again, if necessary via a different path. The most important stumbling block is called head-of-line blocking: a single delayed packet can hold up all subsequent ones in the buffer. Good distributors therefore try not to use very slow paths in the first place.

From Apple's Siri to 5G rollout

The best-known example is Multipath TCP, or MPTCP for short. Apple has been using it for Siri for years, so that voice commands get through even with a shaky WiFi connection. Similar methods are found today in streaming services and in software for corporate networks. The 5G mobile standard also envisions bundling mobile network and WiFi.

In news reports, the term usually appears in connection with network expansion, cloud services, or autonomous vehicles. A self-driving car must not lose its connection to the data center just because a cell tower is missing in a tunnel. Multiple parallel paths are a matter of safety there, not a convenience feature.

A common misconception is that multipath protocols automatically make every connection twice as fast. This only applies if both paths are similarly good and both sides support the method. Often, moreover, it is not about speed at all, but about failover reliability. Anyone who pays close attention notices the difference mainly in the fact that a video call no longer drops when leaving the house.