MHQoS: A Multi-user Hierarchical Quality of Service Optimization for P2P Storage

In peer-to-peer (P2P) storage, particularly during high-concurrency file sharing and massive user access, maintaining robust quality of service (QoS) poses considerable challenges. The existing dmclock scheduling approach in Ceph systems lacks fine-grained prioritization, often resulting in suboptimal performance differentiation under heavy load conditions and a diminished experience for high-priority users. This paper introduces MHQoS, a Multi-user Hierarchical Quality of Service Optimization framework, specifically designed to address these issues. MHQoS enhances resource scheduling accuracy and efficiency through a two-tiered path structure: a microsecond-level issue path for immediate request handling and a millisecond-level planning path for strategic global resource adjustment. By incorporating refined user priority classifications and dynamically adjustable QoS templates, MHQoS achieves precise resource allocation and minimizes latency. Experimental results demonstrate that MHQoS improves high-priority user performance by 51.15%, increases total bandwidth by 5.34%, and reduces latency by 4.09%. Additionally, the framework incorporates a novel work-conserving algorithm that, once each user’s SLO is fulfilled, prioritizes the allocation of remaining system resources to address the needs of high-priority users, thereby further enhancing their overall experience. These enhancements position MHQoS as a accurate, efficient, and scalable solution for dynamic QoS management in P2P storage environments.