Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Understanding Layered Portability from HPC to Cloud in Containerized Environments

Daniel Medeiros, Gabin Schieffer, Jacob Wahlgren, Ivy Peng

TL;DR

This work uses six representative HPC applications to characterize the impact of container virtualization, host OS and kernel, and rootless and privileged container execution, and shows that changing between the container execution modes results in negligible performance differences in the six applications.

Abstract

Recent development in lightweight OS-level virtualization, containers, provides a potential solution for running HPC applications on the cloud platform. In this work, we focus on the impact of different layers in a containerized environment when migrating HPC containers from a dedicated HPC system to a cloud platform. On three ARM-based platforms, including the latest Nvidia Grace CPU, we use six representative HPC applications to characterize the impact of container virtualization, host OS and kernel, and rootless and privileged container execution. Our results indicate less than 4\% container overhead in DGEMM, miniMD, and XSBench, but 8\%-10\% overhead in FFT, HPCG, and Hypre. We also show that changing between the container execution modes results in negligible performance differences in the six applications.

Understanding Layered Portability from HPC to Cloud in Containerized Environments

TL;DR

This work uses six representative HPC applications to characterize the impact of container virtualization, host OS and kernel, and rootless and privileged container execution, and shows that changing between the container execution modes results in negligible performance differences in the six applications.

Abstract

Recent development in lightweight OS-level virtualization, containers, provides a potential solution for running HPC applications on the cloud platform. In this work, we focus on the impact of different layers in a containerized environment when migrating HPC containers from a dedicated HPC system to a cloud platform. On three ARM-based platforms, including the latest Nvidia Grace CPU, we use six representative HPC applications to characterize the impact of container virtualization, host OS and kernel, and rootless and privileged container execution. Our results indicate less than 4\% container overhead in DGEMM, miniMD, and XSBench, but 8\%-10\% overhead in FFT, HPCG, and Hypre. We also show that changing between the container execution modes results in negligible performance differences in the six applications.
Paper Structure (10 sections, 8 figures, 2 tables)

This paper contains 10 sections, 8 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Background
  3. Methodology
  4. Applications
  5. Evaluation
  6. How much impact does the usage of a containerized environment for application deployment cause?
  7. How much does using a container base image different from the host image affect the application?
  8. How much a security-hardened rootless containers affect the performance of applications?
  9. Related Works
  10. Conclusions

Figures (8)

  • Figure 1: A hierarchical view of different layers in a containerized environment
  • Figure 2: The main stages of building a image on a platform and deploying it on another platform.
  • Figure 3: Results for the six HPC applications running on three generations of ARM processor, including Nvidia Grace, ARM Ampere Altra, and APM X-GENE.
  • Figure 4: The relative changes in four performance counters in six HPC applications on the Sleipner platform (Nvidia Grace) in Docker container and Bare Metal.
  • Figure 5: The overhead of Docker container deployment on the Nvidia Grace-based platform in the six applications.
  • ...and 3 more figures