Maxwell must be placed so that the packets to be impaired have to cross through Maxwell on their way from source to destination.

Maxwell acts as a layer 2 bridge that connects two ethernets together.

Maxwell's bridging system is quite simple: packets that arrive on interface 0 are passed through the impairment system and transmitted from interface 1.  Symmetrically packets that arrive on interface 1 are passed through the impairment system and transmitted from interface 0.

Maxwell does not emit IEEE 802 spanning tree packets, but it will relay such packets that are created by other devices.

Because Maxwell does not participate in spanning tree protocols it is important that Maxwell by physically wired into the network in a way that does not create loops.

Usually the easiest and most effective place to put Maxwell is so that it has one interface plugged into your lab network (on which one of the devices under test is located) and the other onto an isolated stub network on which the other devices under test are located.  (The network services that we always forget about, such as DNS and DHCP, would generally be provided from the lab work - Maxwell by default will convey these packets to the devices on the isolated stub network.)

Avoiding Loops

A packet loop will be created if Maxwell is hooked up in a way that allows a packet leaving one of the two Maxwell interfaces to flow around and come back into the other interface.

This usually happens when the user plugs both Maxwell's interfaces into the same switch or into different switches that, in turn, connect to the ethernet same broadcast domain or switch.

Packets will flow around the loop as fast as Maxwell can forward them - this is usually fast enough to significantly annoy other users of that network - and it often makes Maxwell's graphical interface sluggish or non-responsive.

If you create a loop, you should disconnect the cable plugged into either one of Maxwell's interfaces.  (Take care, however, not to unplug the cable from the control interface.)