Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Exploring sustainable alternatives for the deployment of microservices architectures in the cloud

Vittorio Cortellessa, Daniele Di Pompeo, Michele Tucci

TL;DR

A novel approach to support cloud deployment of microservices architectures by targeting optimal combinations of application performance, deployment costs, and power consumption is introduced by leveraging genetic algorithms, specifically NSGA-II.

Abstract

As organizations increasingly migrate their applications to the cloud, the optimization of microservices architectures becomes imperative for achieving sustainability goals. Nonetheless, sustainable deployments may increase costs and deteriorate performance, thus the identification of optimal tradeoffs among these conflicting requirements is a key objective not easy to achieve. This paper introduces a novel approach to support cloud deployment of microservices architectures by targeting optimal combinations of application performance, deployment costs, and power consumption. By leveraging genetic algorithms, specifically NSGA-II, we automate the generation of alternative architectural deployments. The results demonstrate the potential of our approach through a comprehensive assessment of the Train Ticket case study.

Exploring sustainable alternatives for the deployment of microservices architectures in the cloud

TL;DR

A novel approach to support cloud deployment of microservices architectures by targeting optimal combinations of application performance, deployment costs, and power consumption is introduced by leveraging genetic algorithms, specifically NSGA-II.

Abstract

As organizations increasingly migrate their applications to the cloud, the optimization of microservices architectures becomes imperative for achieving sustainability goals. Nonetheless, sustainable deployments may increase costs and deteriorate performance, thus the identification of optimal tradeoffs among these conflicting requirements is a key objective not easy to achieve. This paper introduces a novel approach to support cloud deployment of microservices architectures by targeting optimal combinations of application performance, deployment costs, and power consumption. By leveraging genetic algorithms, specifically NSGA-II, we automate the generation of alternative architectural deployments. The results demonstrate the potential of our approach through a comprehensive assessment of the Train Ticket case study.
Paper Structure (32 sections, 9 equations, 5 figures, 4 tables)

This paper contains 32 sections, 9 equations, 5 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Approach
  3. Objectives
  4. Power consumption
  5. Response Time
  6. Complexity
  7. Cost
  8. Modeling assumptions
  9. Refactoring catalog
  10. REDO - Redeploy Existing Component
  11. MOVE - Relocate Operation to Existing Component
  12. CLON - Clone Node
  13. MOTN - Move Operation to New Component on New Node
  14. DROP - Remove Node
  15. Experimental setup
  16. ...and 17 more sections

Figures (5)

  • Figure 1: UML diagrams
  • Figure 2: Comparison of the Pareto fronts resulting from the baseline (without the power objective, 13 solutions) and the power-aware (with the power objective, 68 solutions) experiments. In the baseline experiment, power consumption was not considered as an optimization objective, but only computed afterward on the models that are part of the Pareto front.
  • Figure 3: Distributions of the objective values in the baseline and the power-aware experiments. Plotted solutions are all the Pareto fronts obtained in the 31 runs.
  • Figure 4: Simplified view of the transformation from UML to LQN.
  • Figure 5: Power consumption and cost of individual types of requests for the solutions in the Pareto front of the power-aware experiment. Entire bars represent the total power consumption and cost of the system for a given solution. The x-axis lists solutions IDs, highlighted in bold if mentioned in the text.