It has nothing to do with how many ports one adapter has. A "path" is just a way of looking at one disk. The short answer is, a "path" is a certain connection from the system to the disk, using a specific way.

Here is the long version: consider the following setup. You have one FC-Adapter (two ports, FCa, FCb), which is connected to two FC-switches (SWa, SWb), which both connect to a storage system. The switches are themselves connected via a crossover-link. You create a single disk on this storage system. This disk will be seen over different paths: FCa-SWa-Disk, FCa-SWa-SWb-Disk, FCb-SWb-Disk, FCb-SWb-SWa-Disk. AIX will therefore create 4 different hdisk-devices when you run cfgmgr.

Now, this is not very comfortable, especially when you create several disks. Each disk will be seen via 4 paths (if you have a fabric more complex than the one i sketched above you might have even more paths) and soon you will have trouble to remember that hdisk137 is in fact hdisk49 and also hdisk87 and hdisk3.

This is where MPIO comes into play: it takes all the hdisk-devices which represent in fact one and only one disk (its like taking a picture of a building from several angles - regardless of how many pictures you take it is always the same one building) and builds a "super-device" above them. You can use this super-device to build VGs, LVs and all other sorts of things you use normally disks for. The MPIO (Multipath-I/O) driver will take care about the relationship of the different hdisk devices and do the bookkeeping for you (and the OS as well).

And there is an additional advantage to it: suppose Switch A in the above picture would fail. Two of the four hdisks (the two paths involving SWa) would be inaccessible, because these pathes would not be usable, but the super-device still works, because it uses ALL paths and if two fail there are still two more to use. In fact this is the main reason one builds such complex fabrics with parallel FC-switches, crossover-links and the like.

I hope this helps.

bakunin