MAPA: Multi-Accelerator Pattern Allocation Policy for Multi-Tenant GPU Servers
MAPA: Multi-Accelerator Pattern Allocation Policy for Multi-Tenant GPU Servers
复制标题
DOI：
10.1145/3458817.3480853
复制
发表时间：
2021-10
期刊：
影响因子：
--
通讯作者：
K. Ranganath;Joshua D. Suetterlein;J. Manzano;S. Song;Daniel Wong
中科院分区：
文献类型：
--
作者：
K. Ranganath;Joshua D. Suetterlein;J. Manzano;S. Song;Daniel Wong
多加速器服务器越来越多地部署在共享的多租户环境中（例如云数据中心），以满足大规模计算密集型工作负载的需求。此外，这些加速器越来越多地在复杂的拓扑结构中相互连接，工作负载显示出各种各样的加速器间通信模式。然而，现有的分配策略并不适合这些新出现的用例。具体来说，这项工作确定了多加速器工作负载通常是碎片化的，导致加速器间通信的带宽减少和延迟增加。我们提出了多加速器模式分配（MAPA），这是一种图形模式挖掘方法，旨在为在多加速器服务器上分配多加速器工作负载提供通用的分配支持。我们证明了MAPA能够改善多加速器工作负载的执行时间，并且MAPA能够跨各种加速器拓扑提供普遍的好处。最后，我们演示了在使用MAPA的基准策略下，对于第75百分位的作业，加速提高了12.4%，最坏情况下的执行时间最多减少了35%。
Multi-accelerator servers are increasingly being deployed in shared multi-tenant environments (such as in cloud data centers) in order to meet the demands of large-scale compute-intensive workloads. In addition, these accelerators are increasingly being inter-connected in complex topologies and workloads are exhibiting a wider variety of inter-accelerator communication patterns. However, existing allocation policies are ill-suited for these emerging use-cases. Specifically, this work identifies that multi-accelerator workloads are commonly fragmented leading to reduced bandwidth and increased latency for inter-accelerator communication. We propose Multi-Accelerator Pattern Allocation (MAPA), a graph pattern mining approach towards providing generalized allocation support for allocating multi-accelerator workloads on multi-accelerator servers. We demonstrate that MAPA is able to improve the execution time of multi-accelerator workloads and that MAPA is able to provide generalized benefits across various accelerator topologies. Finally, we demonstrate a speedup of 12.4% for 75th percentile of jobs with the worst case execution time reduced by up to 35% against baseline policy using MAPA.