Comparison Study about Heterogeneous Size Estimation Methods

Modern systems appear large, distributed and heterogeneous. It is often required to know the number of certain kinds of items for optimization. Such problem is called Size Estimation Problem. Until now, researchers have proposed quite a lot of size estimation methods based on sampling methods which using limited observing data to reckon the total size. In this article, we focus on comparison of these methods in various heterogeneous environments by experimental simulations. We find that these methods lead to serious system errors in some heterogeneous environments. And we also find it is impossible to take any improving measures in unknown heterogeneous situations, because the observing data are same in different heterogeneous systems.