一句话：提出 FleetIO 学习型存储虚拟化框架，用多智能体强化学习动态调度多租户 SSD 共享，兼顾性能隔离与利用率。
云平台通过虚拟化 SSD 等存储设备实现多租户存储共享，但长期面临性能隔离与资源利用率之间的根本矛盾：弱隔离导致长尾延迟，强隔离导致存储利用率低下。本文提出 FleetIO，一种基于学习的存储虚拟化框架，使用强化学习管理虚拟化 SSD。FleetIO 利用 RL 应对应用工作负载和存储状态的动态变化，将存储调度集成到 RL 决策过程中，通过在同驻应用实例间进行动态细粒度存储收割，在尽量不影响服务级目标（SLO）的前提下，同时实现性能隔离与更高存储利用率。它根据运行时采集的 I/O trace 将工作负载聚类为不同类型（如延迟敏感型和带宽密集型），并针对每类工作负载微调 RL 奖励函数。作者在真实可编程 SSD 板上实现 FleetIO，并使用多种云应用进行评估。结果表明，与最先进的存储共享方法相比，FleetIO 将共享 SSD 的整体存储利用率最高提升 1.4 倍，并将 I/O 请求尾延迟平均降低 1.5 倍。
Cloud platforms have been virtualizing storage devices like flash-based solid-state drives (SSDs) to make effective use of storage resources. They enable either software-isolated instance or hardware-isolated instance for facilitating the storage sharing between multi-tenant applications. However, for decades, they have to combat the fundamental tussle between the performance isolation and resource utilization. They suffer from either long tail latency caused by weak isolation or low storage utilization caused by strong isolation. In this paper, we present FleetIO, a learning-based storage virtualization framework that employs reinforcement learning (RL) for managing virtualized SSDs. FleetIO explores the unique features of RL to handle the dynamic changes of application workloads and storage states, and integrates the storage scheduling into the RL decision-making process. It achieves both performance isolation and improved storage utilization by enabling dynamic fine-grained storage harvesting across collocated application instances, while minimizing its negative impact on their service-level objectives (SLOs). FleetIO clusters workloads into different types (e.g., latency-sensitive and bandwidth-intensive) based on the collected I/O traces at runtime, and fine-tunes the RL reward functions for each type of workloads. We implement FleetIO on a real programmable SSD board and evaluate it with diverse cloud applications. We show that FleetIO improves the overall storage utilization of the shared SSD by up to 1.4×, and decreases the tail latency of I/O requests by 1.5× on average, compared to the state-of-the-art storage sharing approaches.
@inproceedings{SunRLCSH25,
title = {FleetIO: Managing Multi-Tenant Cloud Storage with Multi-Agent Reinforcement Learning},
author = {Jinghan Sun and Benjamin Reidys and Daixuan Li and Jichuan Chang and Marc Snir and Jian Huang},
booktitle = {ASPLOS},
year = {2025},
doi = {10.1145/3669940.3707229}
}