Regulating I/O Performance of Shared Storage with a Control Theoretical Approach.
Han Deok Lee, Young Jin Nam, Kyong Jo Jung, Seok Gan Jung, Chanik Park · 2004
Shared storage has become commonplace with recent trends in storage technologies, such as storage consolidation and virtualization, etc. Meanwhile, storage QoS, which guarantees different storage service requirements from various applications toward shared storage, is gaining in importance. This paper proposes a new scheme which combines a feedback-controlled leaky bucket with a fair queuing algorithm in order to deliver guaranteed storage service for applications competing for shared storage. It not only assures an agreed-upon response time for each application, but also maximizes the aggregate I/O throughput by proportionating unused bandwidth to other active applications. Simulation results under various types of competing I/O workloads validate the features of the proposed scheme.