QoS Guarantees and Performance Isolation for Multi-tenant SSDs in Cloud Environments

Abstract

With the widespread adoption of cloud computing and virtualization, multi-tenant architecture has become the mainstream deployment model for data center storage. Solid-state drives (SSDs), leveraging high IOPS, low latency, and high parallelism, serve as the core storage medium. Nevertheless, internal resource contention, garbage collection (GC) interference and heterogeneous workloads constantly induce cross-tenant I/O conflicts, unequal performance allocation and latency fluctuation. These issues greatly weaken the predictability of storage quality of service (QoS) and form a critical bottleneck for multi-tenant storage systems. This paper systematically reviews the research background, core challenges, and key technologies of SSD QoS guarantee, focusing on flash-level isolation, software scheduling, and end-to-end collaborative optimization in multi-tenant contexts. It analyzes the strengths and limitations of existing strategies, explores the integration of emerging technologies such as zoned namespace (ZNS) and compute express link (CXL) with SSD QoS guarantee, discusses the gaps in interference modeling and isolation mechanisms, adaptation to novel storage paradigms, and QoS evaluation frameworks, and covers the link between QoS and SSD package-level reliability. This work aims to clarify the research trajectory, provide references for follow-up studies, and facilitate the efficient, large-scale, and reliable deployment of SSDs in cloud data centers, edge computing, and heterogeneous computing environments.