Inverse multiplexing is a communications networking technique by which multiple connection circuits can be used to form a single data line. The technique is sometimes shortened to inverse muxing, or simply IMUX. With such a network connection, each of the smaller circuits is logically grouped, then understood by the entire network as one communication pipeline. IMUX is, therefore, capable of providing a larger amount of overall bandwidth, as the data is spread across the smaller communication circuits.
It can be easy to confuse inverse multiplexing with regular multiplexing. In normal multiplexing, multiple data streams are directed into a large networking circuit and separated back out again at the other end. With an inverse multiplex, the house itself could be included in the move by taking it apart, putting it on several trucks, and rebuilding it at the new location. In this analogy, the IMUX is happening at both ends of the move.
When it comes to networking, there are a couple methods to performing inverse multiplexing, both corresponding to the open systems interconnection (OSI) model's first two layers, the physical layer and the data link layer. At which layer the IMUX takes place depends on the type of network being established, such as frame relay, Ethernet, and so forth. In the case of a layer one method, the physical layer, it is the very bits that are separated out into the different circuits that comprise the connection. In a layer two method, the data link layer, where packets are converted into frames for transport over the physical layer, the frames are distributed amongst the circuits.
The way most inverse multiplexing implementations work is via a round robin approach, where the data is sent to each smaller circuit, one at a time, and starts back again with the first. Some differences occur depending on the implementation, however. With a frame relay method, a larger frame may go out on the first circuit, and by the time it's finished, the other circuits may have sent many more smaller frames. With an asynchronous transfer mode (ATM) network, every cell in an ATM network is the same size, garnering greater efficiency in the distribution process.
Besides the cost-effectiveness of inverse multiplexing, there are other benefits as well, such as more accurate load balancing among the network, and lower latency on any given circuit. The network is also more flexible to deal with growth. As growth justifies implementing higher-speed circuits in one area of a wide area network (WAN), the IMUX hardware can be re-purposed to another area of the WAN.
