Cloud servers are basically developed for supporting the efficient communication and the computation remotely. Thus a number of different time zone machines are able to find the services of resources and the computation. Thus each and every time these machines are working in order to resolve the end client request. But sometime these machines are loaded more then their capacity and does not perform as desired and some of the available processing units are considered to be free. This results in the performance loss in the computational servers. In order to find the optimum performance there is need to implement some technique for load balancing. Thus in this presented work the key focus is placed on the investigation of load balancing approaches and the algorithm. In order to find the optimum technique of load balancing, various different algorithms are studied. Among them four most promising techniques are selected which are promising for load balancing in previous studies. These techniques are genetic algorithm which provides the optimization technique for finding most appropriate solution among available solutions, ABC (artificial honey bee colony) algorithm which is also an optimization technique in the similar ways, the round robin which is frequently used for process allocation during CPU scheduling and finally the self-organizing map which is an unsupervised class of machine learning. These methods are implemented and compared on the basis of their performance parameters. The implementation of the proposed comparative study is performed on the JAVA technology and using the Cloud Sim simulator. After simulation of the presented comparative study the performance of the SOM algorithm provides the optimum results as compared to the Genetic algorithm, Honey bee colony and RR (round robin) algorithms.