Load Balancing in Compute Clusters with Delayed Feedback

Load balancing arises as a fundamental problem, underlying the dimensioning and operation of many computing andcommunication systems, such as job routing in data center clusters, multipath communication, Big Data and queueing systems. Inessence, the decision-making agent maps each arriving job to one of the possibly heterogeneous servers while aiming at anoptimization goal such as load balancing, low average delay or low loss rate. One main difficulty in finding optimal load balancingpolicies here is that the agent only partially observes the impact of its decisions, e.g., through the delayed acknowledgements of theserved jobs. In this paper, we provide a partially observable (PO) model that captures the load balancing decisions in parallel bufferedsystems under limited information of delayed acknowledgements. We present a simulation model for this PO system to find a loadbalancing policy in real-time using a scalable Monte Carlo tree search algorithm. We numerically show that the resulting policyoutperforms other limited information load balancing strategies such as variants of Join-the-Most-Observations and has comparableperformance to full information strategies like: Join-the-Shortest-Queue, Join-the-Shortest-Queue(d) and Shortest-Expected-Delay.Finally, we show that our approach can optimise the real-time parallel processing by using network data provided by Kaggle