Fog Computing (FC) enhances the proficiency and performance of cloud with the objective of bringing selected computing capabilities to the network edge to reduce latency, enhance location awareness, provide mobility support, etc. In the FC environment, resources could not be utilized completely due to the wide variation in execution duration and specifications for computing nodes. Therefore, Load Balancing (LB) for the computing nodes in the FC environment is an essential aspect that avoids the situation of under-or over-loaded fog nodes during the execution of Internet of Things (IoT) applications, especially. This paper starts by presenting several FC architectures, their comparisons and LB strategies in different computing environments. An empirical survey of the existing LB mechanisms has been presented along with trade-offs in their performance metrics. Various possible LB metrics have been investigated. Finally, potential challenges, case studies and future directions on LB in FC environments are presented.