The use of fluid generalized processor sharing (GPS) algorithm for integrated services networks has received a lot of attention since early 1990s because of its desirable properties in terms of delay bound and service fairness. Many packet fair queuing (PFQ) algorithms have been developed to approximate GPS. However, owing to their implementation complexity, it is difficult to support a large number of sessions with diverse service rates while maintaining the GPS properties. The grouping architecture has been proposed to dramatically reduce the implementation complexity. However, it can only support a fixed number of service rates, thus causing the problem of granularity. We present a viable implementation of our previously proposed dual-rate grouping architecture, and demonstrate that, as compared with the original grouping architecture, our proposed scheme possesses better performance in terms of approximating per session-based PFQ algorithms without increasing the implementation complexity.