DILATON: Multi-dimensional Quantum Deficit Algorithm for Multi-tenant SSD Resource Allocation

In modern cloud infrastructures, multi-tenancy and virtualization are common, yet optimizing resource allocation while maintaining performance is challenging due to complex Solid State Device (SSD) physics. Over-provisioning resources often leads to revenue loss and higher infrastructure costs. To address this, we introduce DILATON, an adaptive closed-loop scheduler that balances bandwidth and QoS across tenants. By predicting die loading and using real-time adaptation, DILATON maximizes resource allocation, achieves performance gains, and reduces costs while meeting SLAs. This paper uses simulations to emphasize the necessity of a closed-loop QoS control.