Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Cost-Effective Edge Data Distribution with End-To-End Delay Guarantees in Edge Computing

Ravi Shankar, Aryabartta Sahu

TL;DR

This paper addresses the problem of distributing application data across edge servers to minimize combined cloud-to-edge and edge-to-edge costs while satisfying end-to-end delay constraints. It introduces a refined Integer Programming formulation (EDD) for exact optimization and a scalable Network Steiner Tree Estimation-based heuristic (EDD-NSTE) for large-scale instances, with a theoretical $O(k)$-approximation guarantee. Empirical evaluation on EUA and SLNDC demonstrates that EDD-NSTE closely matches the IP solution and significantly outperforms greedy or random baselines, achieving up to 80.35% cost savings in tested scenarios. The work provides a practical framework for cost-efficient, delay-aware edge data distribution that can guide real-world edge deployment and caching strategies under latency requirements.

Abstract

Cloud Computing is the delivery of computing resources which includes servers, storage, databases, networking, software, analytics, and intelligence over the internet to offer faster innovation, flexible resources, and economies of scale. Since these computing resources are hosted centrally, the data transactions from the cloud to its users can get very expensive. Edge Computing plays a crucial role in minimizing these costs by shifting the data from the cloud to the edge servers located closer to the user's geographical location, thereby providing low-latency app-functionalities to the users of that area. However, the data transaction from the cloud to each of these edge servers can still be expensive both in time and cost. Thus, we need an application data distribution strategy that minimizes these penalities. In this research, we attempt to formulate this Edge Data Distribution as a constrained optimization problem with end-to-end delay guarantees. We then provide an optimal approach to solve this problem using the Integer Programming (IP) technique. Since the IP approach has an exponential time complexity, we also then provide a modified implementation of the EDD-NSTE algorithm, for estimating solutions to large-scale EDD problems. These algorithms are then evaluated on standard real-world datasets named EUA and SLNDC and the result demonstrates that EDD-NSTE significantly outperformed, with a performance margin of 80.35\% over the other representative approaches in comparison.

Cost-Effective Edge Data Distribution with End-To-End Delay Guarantees in Edge Computing

TL;DR

This paper addresses the problem of distributing application data across edge servers to minimize combined cloud-to-edge and edge-to-edge costs while satisfying end-to-end delay constraints. It introduces a refined Integer Programming formulation (EDD) for exact optimization and a scalable Network Steiner Tree Estimation-based heuristic (EDD-NSTE) for large-scale instances, with a theoretical -approximation guarantee. Empirical evaluation on EUA and SLNDC demonstrates that EDD-NSTE closely matches the IP solution and significantly outperforms greedy or random baselines, achieving up to 80.35% cost savings in tested scenarios. The work provides a practical framework for cost-efficient, delay-aware edge data distribution that can guide real-world edge deployment and caching strategies under latency requirements.

Abstract

Cloud Computing is the delivery of computing resources which includes servers, storage, databases, networking, software, analytics, and intelligence over the internet to offer faster innovation, flexible resources, and economies of scale. Since these computing resources are hosted centrally, the data transactions from the cloud to its users can get very expensive. Edge Computing plays a crucial role in minimizing these costs by shifting the data from the cloud to the edge servers located closer to the user's geographical location, thereby providing low-latency app-functionalities to the users of that area. However, the data transaction from the cloud to each of these edge servers can still be expensive both in time and cost. Thus, we need an application data distribution strategy that minimizes these penalities. In this research, we attempt to formulate this Edge Data Distribution as a constrained optimization problem with end-to-end delay guarantees. We then provide an optimal approach to solve this problem using the Integer Programming (IP) technique. Since the IP approach has an exponential time complexity, we also then provide a modified implementation of the EDD-NSTE algorithm, for estimating solutions to large-scale EDD problems. These algorithms are then evaluated on standard real-world datasets named EUA and SLNDC and the result demonstrates that EDD-NSTE significantly outperformed, with a performance margin of 80.35\% over the other representative approaches in comparison.
Paper Structure (15 sections, 31 equations, 21 figures, 1 table, 3 algorithms)

This paper contains 15 sections, 31 equations, 21 figures, 1 table, 3 algorithms.

Table of Contents

  1. Introduction
  2. Related Work
  3. Problem Formulation
  4. Solution Strategy
  5. Edge Data Distribution (EDD) as an Integer Programming Problem
  6. Example to understand the Integer Programming Approach
  7. Edge Data Distribution (EDD) estimation as a Network Steiner Tree Estimation Problem
  8. Example to understand the EDD-NSTE algorithm
  9. Time Complexity analysis of EDD-NSTE algorithm
  10. EDD-NSTE is an O(k) approximation algorithm
  11. Experimental Evaluation
  12. Simulation Settings and Approaches for comparison of the result
  13. Dataset Used
  14. Experiment Results
  15. Conclusion

Figures (21)

  • Figure 1: EDD scenario with 10 edge servers
  • Figure 2: EDD example to demonstrate Llimit
  • Figure 3: Optimal Solution using integer programming
  • Figure 4: Steiner Tree estimated by Algorithm 1
  • Figure 5: Final EDD solution estimated by Algorithm 3
  • ...and 16 more figures