A common approach to deal with NP-hard problems is to deploy polynomial-time ϵ-approximation algorithms. These algorithms often resort to rounding and scaling to guarantee a solution that is within a factor (1 + isin) of the optimal solution. Usually, researchers either only round up or only down. In this paper we will evaluate the gain in accuracy when rounding up and down. The main application of this technique upon which we focus is quality of service routing, and specifically the restricted shortest path problem