Although using drop and insert enables bandwidth calculator to be shared, the bandwidth cannot be allocated dynamically. Thus, one of the advantages of inverse multiplexing--bandwidth on demand--is not made available. The other advantage--grouping low-bandwidth channels into larger-bandwidth circuits--is still viable, however.

You can use inverse multiplexing at the packet level or at the circuit level. At the packet level, packets in a data stream are allocated to the available channels supported by the I-Mux. For example, an inverse multiplexer that had selected three Switched 56 channels would send the first packet it received to channel 1, the second to channel 2, the third to channel 3, the fourth to channel 1 again, and so forth. At the receiving end, an I-Mux reassembles the packets into a coherent data stream. When packet-level inverse multiplexing is used, LAN packets are encapsulated into WAN packets using a standard protocol such as X.25, the Point-to-Point Protocol (PPP), or the I-Mux manufacturer's proprietary WAN protocol.

On the other hand, if the inverse multiplexer uses circuit-level switching, it subdivides the data stream into equal portions without regard to data or packets and sends each portion over one of the channels. At the receiving end, the sections of the data streams are, again, reassembled. Packet-level inverse multiplexing is sometimes called load balancing, since it effectively divides the entire data stream at the packet level, resulting in relatively equal traffic on each channel. Circuit switching, however, does not work at the packet level, but on a time basis. Thus, a given portion of the data stream may contain more actual data than another portion.

I-MUX as a Router If you are going to create virtual circuits using inverse multiplexing, you will need a method of routing data from the source to its intended destination. An I-Mux access device can act as a router in several ways. The first and easiest technique is called static routing. With this method, the routes that packets will take are fixed between source and destination. Since there is no discovery involved in static routing, maintaining inter network connections using switched lines poses no difficulty. However, you may encounter some difficulty when using other routing techniques, such as Routing Information Protocol (RIP) or Open Shortest Path First (OSPF).

The best approach is to view a switched circuit as a dedicated circuit that is either up or down. If it is up, RIP or OSPF can do their work, routing such protocols as IP, IPX, AppleTalk, and so forth. If the switched line is down, the router/I-Mux views the circuit as down and responds accordingly. You can limit overhead traffic by turning RIP off.

For the I-Mux to make a connection over a WAN, it must establish a call. First, the I-Mux must know that it is to make a connection and it must know the destination. Second, it must know how to get its packets to that destination. Most inverse multiplexers can establish a call manually, based on the time of day, the date, or the destination address of a packet. Once an initial connection has been established, the I-Mux can make additional calls--and add additional bandwidth--based on fixed parameters such as the time of day or the traffic load.