The large amount of data processing is growing fast. In many scenarios, both in industry and academia, high-performance computing (HPC) is a needed to process Big Data in short time. An important topic in HPC is task scheduling. In this context, several scheduling algorithms for grid computing have been proposed, but these algorithms have not been yet fully tested by taking into account different types of applications and platforms. This paper presents a comparative study designed to discover the behavior of scheduling algorithms exhibit under different settings. This was done through a methodology with four criteria: performance, scalability, workload distribution and adaptability. The scheduling algorithms compared in this paper are HEFT, CPOP and PCH. Our results show that the HEFT algorithm performs well in almost all cases, despite its simplicity. The PCH and CPOP algorithms perform well only in very specific cases, this is mainly due to the high dependence on the critical path for both algorithms.