Model Predictive Control of Internet Servers using Quantum Annealing
Efthymios Papageorgiou, George T. Stamatiou, Kostas Magoutis · 2024
Internet servers must adapt rapidly to workload or configuration changes. Optimal control-based approaches like Model Predictive Control (MPC) offer proactive adaptation but present challenges, including the need to take into account reconfiguration features of modern server architectures and the computation cost of the MPC optimization problem. This work examines these challenges using the Apache HTTP Server’s graceful restart for dynamic reconfiguration. We solve the MPC optimization problem using both classical and quantum optimization solutions, comparing their effectiveness. Evaluating our MPC implementation on Apache HTTP Server with classical and quantum annealing-based (QA) D-Wave quantum hardware, our results show that QA-based solutions lead to an effective MPC controller, similar to classical optimization with appropriate tuning. As problem complexity and integration with quantum computers increase, QA-based solutions are expected to outperform classical alternatives for dynamic control.