<p>Polymorphic packet scheduler refers to schedule diverse traffic with different Quality of Service (QoS) requirements in the multi-tenant cloud environment. However, existing proposals either cannot achieve traffic management between multi-tenant and multi-service or cannot support customized QoS requirements. We propose a packet scheduler structure that supports polymorphic scheduling algorithms and Hierarchical QoS (HQoS), called Polymorphic HQoS (Poly-HQoS), which can support fine-grained traffic management between different tenants and services and provide customized scheduling algorithms. We design an adaptive algorithm to implement customized queue scheduling in the first level and weighted Max-Min fair scheduling in the second level based on the hierarchical rank distribution and admission control. Finally, we implement the Poly-HQoS prototype on hardware Barefoot Tofino switches and a large-scale simulator, demonstrating its ability to ensure traffic isolation and coexistence of different flow aggregations while supporting customized programmable queue scheduling algorithms and achieving better performance by reducing the average flow completion time for short flows up to 70%.</p>

错误:搜索内容不能为空,请输入英文关键词
错误:关键词超出字数限制,请精简
高级检索

Poly-HQoS: a polymorphic packet scheduler for traffic isolation in multi-tenant cloud environment

  • Saifeng Hou,
  • Yuxiang Hu,
  • Le Tian,
  • Pengshuai Cui

摘要

Polymorphic packet scheduler refers to schedule diverse traffic with different Quality of Service (QoS) requirements in the multi-tenant cloud environment. However, existing proposals either cannot achieve traffic management between multi-tenant and multi-service or cannot support customized QoS requirements. We propose a packet scheduler structure that supports polymorphic scheduling algorithms and Hierarchical QoS (HQoS), called Polymorphic HQoS (Poly-HQoS), which can support fine-grained traffic management between different tenants and services and provide customized scheduling algorithms. We design an adaptive algorithm to implement customized queue scheduling in the first level and weighted Max-Min fair scheduling in the second level based on the hierarchical rank distribution and admission control. Finally, we implement the Poly-HQoS prototype on hardware Barefoot Tofino switches and a large-scale simulator, demonstrating its ability to ensure traffic isolation and coexistence of different flow aggregations while supporting customized programmable queue scheduling algorithms and achieving better performance by reducing the average flow completion time for short flows up to 70%.