Q-Learning-Based Task Allocation for High-Speed Distributed Computing in Heterogeneous Clusters

This paper proposes a Q-learning-based task allocation approach for wireless coded distributed computing systems with heterogeneous worker nodes. Task allocation in such systems is challenging due to the heterogeneity in computation and communication capabilities, leading to non-identically and independently distributed processing times across nodes. By modeling the task allocation problem as a Markov decision process and applying Q-learning, the master node learns to allocate tasks effectively, adapting to node heterogeneity and minimizing the average processing time. This approach highlights the potential of reinforcement learning to optimize distributed computing in heterogeneous environments.