In Federated Learning (FL), the heterogeneity of participating clients in terms of data distribution, hardware capabilities, and network connectivity introduces significant trade-offs and optimization challenges. These challenges include balancing model accuracy against the constraints of computational resources, efficient training time against potential compromises in model reliability, and fair participation across a diverse range of devices with varying data quality and network capabilities. Managing these complexities require effective client selection strategies to optimize both the performance and efficiency of FL systems. In this chapter, we highlight the critical role of client selection in FL and provide an overview of associated challenges, including heterogeneity, privacy issues, fairness and contextual considerations. The chapter introduces a taxonomy of client selection strategies, offering a structured overview and comparative analysis to evaluate their effectiveness and limitations. Through a synthesis of existing research, we explore strategic approaches to optimize FL system performance and conclude the chapter by discussing implications of our findings for future research, emphasizing the development of more sophisticated and contextually adaptive client selection strategies.