Grid computing is a giant leap forward in high performance distributed computing, ushering in a new era of commodity computing. The biggest challenge in this field is the heterogeneity of applications, resources and administrative domains controlling them. New algorithms of job scheduling and load balancing require stringent evaluation involving controlled configurations. Such a controlled environment is very difficult to provide in a real grid environment. Grid simulators can simulate real grid systems as well as provide configurable environments for repetitive experiments. GridSim is a popular grid simulator originally proposed to evaluate resource management and job scheduling schemes. In this paper we propose an enhancement in GridSim architecture "Resource Topology Aware (RTA) GridSim" to increase its potential in resource management, job scheduling and load balancing. Our measured results show that the proposed enhancement improves the flexibility of GridSim by accommodating resource heterogeneity. In addition, it minimizes communication overhead and response time by making load balancing decisions locally through resource topology awareness.