Hardware virtualization, employing virtual machines (VMs) and/or containers, has gained popularity for deploying systems. Among the container-based solutions, the hybrid approach, where containers run on VMs, has garnered attention from cloud computing providers due to its combination of flexibility and isolation. However, this hybrid setup introduces additional overhead to the system, impacting various components that must be carefully considered during system analysis and dimensioning. This paper evaluates the performance of a real system (SIGA) implemented in two virtualized scenarios: one hosting the application on VMs in an environment with connectivity, storage, and processing; the other running the application in containers hosted on VMs, adding an extra virtualization layer. We assess the number of served requests, CPU, memory, and disk usage as well as the network activity considering the various system elements under critical load. The results reveal that in the scenario with only VMs, 95.5% of requests were successfully served, with an average response time of 2679 ms. In the second scenario, 95.6% of requests were served with an average response time of 2215 ms and lower memory consumption, although CPU and network usage remained similar.

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

Performance Evaluation of Hybrid Virtualization System Deployment: A Case Study of SIGA Application

  • Caio Souza,
  • Israel Duarte,
  • Andson Balieiro,
  • Jamilson Dantas,
  • Elton Alves

摘要

Hardware virtualization, employing virtual machines (VMs) and/or containers, has gained popularity for deploying systems. Among the container-based solutions, the hybrid approach, where containers run on VMs, has garnered attention from cloud computing providers due to its combination of flexibility and isolation. However, this hybrid setup introduces additional overhead to the system, impacting various components that must be carefully considered during system analysis and dimensioning. This paper evaluates the performance of a real system (SIGA) implemented in two virtualized scenarios: one hosting the application on VMs in an environment with connectivity, storage, and processing; the other running the application in containers hosted on VMs, adding an extra virtualization layer. We assess the number of served requests, CPU, memory, and disk usage as well as the network activity considering the various system elements under critical load. The results reveal that in the scenario with only VMs, 95.5% of requests were successfully served, with an average response time of 2679 ms. In the second scenario, 95.6% of requests were served with an average response time of 2215 ms and lower memory consumption, although CPU and network usage remained similar.