Cell breathing is a mechanism used in cellular networks since many years, which is now proposed for energy efficient purposes. By using this technique, it is possible to save energy during low traffic periods by switching unnecessary BSs off whereas the cell topology is re-adapted supported on some few BSs active. In this article, we analyze a multi-size approach which provide means for an enhanced cell topology adaptation. We divide our approach in two redistribution phases. Each of these two phases is conceived for being efficient at specific traffic levels. A first phase configuration only allows short range cell expansion to cover just direct adjacent sectors, which works well for medium traffic load. Then if low load conditions afford it, a second phase may be executed which allows an extended multi-size expansion making it possible to cover additional sectors at a far distance, which results in some new larger cells and some extra BSs switched-off. The decision mechanism makes it possible to choose the best strategy between running "phase 1 only" or "phase 1 & 2 combined". Our approach is supported on a clustered architecture which distributes the computational load in the network. Each cluster consists of groups of BSs with a BS cluster head in charge of making decisions based on collected power saving indicators. Simulation and results are provided to analyze each of the potential approaches supported on phase 1 and 2 redistribution. It is shown that some interesting energy savings are obtained when each of our proposed phases is conveniently applied.