Abstract: The present disclosure provides a framework (120) and a method of optimizing load balancing in a cluster by effectively migrating docker containers which leverages the container runtime real-time metrics of the computing host machine in the virtual environment to decide the optimal container to node placement in the cluster. The present disclosure uses fly container migration in which migrating a container involves check-pointing it on the host node and restoring it on the target node, it also enables synchronizing the container file system, by leveraging the use of private container registry. Load balancing of the instant disclosure is performed by optimized allocation of nodes in the applications thereby re-balancing the cluster and further the framework is scalable to large number of nodes and it leverages the power of graphical processing units (GPU) for making faster scheduling decisions.