As an advanced CFD approach, LES is very promising for practi-cal applications. LES has been successfully applied to many industrial problems, in contrast to DNS, which has been mainly restricted to simple physical problems to understand the ow physics. LES has to be time-dependent three-dimensional calculations, similar to DNS. LES oers several advantages over the traditional CFD based on the RANS modeling approach and DNS. LES is able to predict instantaneous ow characteristics and resolve the large turbulent ow structures, in contrast to the traditional CFD based on the RANS approach, which provides only averaged ow quantities. Compared with DNS, the small scales are modeled in a LES approach and the requirement on accuracy of the numerical schemes is not as high as that of DNS. LES can be applied to problems with complex geometry using unstructured mesh and nite volume methods, but current DNS deals with only simple geometries using predominantly nite dierence and spectral methods to achieve high-order accuracy. In terms of computational costs, LES is also much cheaper than DNS since very ne mesh is not required to resolve the small scales as in DNS.