Water filling may not good neighbors make

Consider a wireless system with multiple and independent user-base pairings over some region. Each user tries to greedily optimize its performance, and eventually a simultaneous water-filling fixed point is reached. Here we seek to analytically understand properties of different water-filling fixed points. In so doing we show that water-filling alone does not generally result in optimum resource sharing and in some cases is a poor solution. We close by suggesting dynamic strategies for performance enhancement.