FairGV: Fair and Fast GPU Virtualization
FairGV: Fair and Fast GPU Virtualization
复制标题
DOI：
10.1109/tpds.2017.2717908
复制
发表时间：
2017-12
影响因子：
5.3
通讯作者：
Cheol-Ho Hong;I. Spence;Dimitrios S. Nikolopoulos
中科院分区：
文献类型：
--
作者：
Cheol-Ho Hong;I. Spence;Dimitrios S. Nikolopoulos
越来越多的高性能计算（HPC）应用程序开发人员由于更高的可用性而选择使用云资源。对于使用云托管服务的HPC应用程序开发人员来说，虚拟化GPU将是一个明显且有吸引力的选择。遗憾的是，现有的GPU虚拟化软件还没有准备好解决与整合混合HPC工作负载相关的公平性、利用率和性能限制。本文介绍了FairGV，这是一种经过彻底重新设计的GPU虚拟化系统，可在使用不同强度GPU的混合工作负载中实现系统范围的加权公平共享和强大的性能隔离。为了实现其目标，FairGV引入了一个无陷阱的GPU处理架构，一个新的公平排队方法与工作保存和以GPU为中心的协同调度策略相结合，以及一个非抢占式GPU的协同调度方法。我们的原型实现实现接近理想的公平性（$\geq 0.97$最小最大比），在一系列利用GPU的混合HPC工作负载的性能下降（$\leq 1.02$聚合开销）。
Increasingly high performance computing (HPC) application developers are opting to use cloud resources due to higher availability. Virtualized GPUs would be an obvious and attractive option for HPC application developers using cloud hosting services. Unfortunately, existing GPU virtualization software is not ready to address fairness, utilization, and performance limitations associated with consolidating mixed HPC workloads. This paper presents FairGV, a radically redesigned GPU virtualization system that achieves system-wide weighted fair sharing and strong performance isolation in mixed workloads that use GPUs with variable degrees of intensity. To achieve its objectives, FairGV introduces a trap-less GPU processing architecture, a new fair queuing method integrated with work-conserving and GPU-centric coscheduling polices, and a collaborative scheduling method for non-preemptive GPUs. Our prototype implementation achieves near ideal fairness ($\geq 0.97$ Min-Max Ratio) with little performance degradation ($\leq 1.02$ aggregated overhead) in a range of mixed HPC workloads that leverage GPUs.