Basically when fatigue tests are done, they are done on a rotating fixture with a weight on it. As the test sample rotates, the loads at at any given point are cycled from compression to tension depending on where that point is in its path of rotation. The load is set at the beginning of the test and the sample is run until failure. Then another load is put on another test sample and the process is repeated until you can make a graph from the data. Thus when you look at a graph of fatigue life, you are looking at cycles to failure at a given load.

In the case of steel, this graph will level off (or nearly so) at some value for load, and this is called the fatigue strength of the material. Steel that is loaded below its fatigue strength is assumed to be able to withstand an infinite number of cycles.

Aluminum, on the other hand, never levels off. This is why it is said to have "no fatigue strength". Thus, even with very low loading it will still eventually fail, albeit after a very large number of cycles. Part of this is due to the fact that aluminum cracks as it is cycled, and with every cycle the cracks grow a little more. A good example of this is around the rivet holes on an airplane (which is where the stress concentration is highest). Any aluminum-skinned airplane flying today will this cracking to some extent,and the airlines are actually required to track the length of the cracks so that they know when to perform a repair. This is also why airliners are no longer eligable for service in the US after a certain number of pressurization cycles.