Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

PSMOA: Policy Support Multi-Objective Optimization Algorithm for Decentralized Data Replication

Xi Wang, Susmit Shannigrahi

TL;DR

PSMOA addresses the challenge of policy-aware, multi-objective data replication in decentralized, multi-organizational environments by integrating NSGA-III with a dynamic policy framework. It uses Entropy Weighted TOPSIS to assign objective weights and generate reference points, while a policy expression module adapts weights in response to system conditions and user constraints. The approach optimizes four objectives—replication time, cost, popularity, and load balance—under storage and bandwidth constraints, leveraging NSGA-III to maintain solution diversity and convergence. Experimental results show PSMOA achieving superior generational distance and inverted generational distance compared to NSGA-II/III, along with notable load-balancing gains and effective dynamic policy adaptation, indicating strong practical potential for federated storage systems like the WLCG.

Abstract

Efficient data replication in decentralized storage systems must account for diverse policies, especially in multi-organizational, data-intensive environments. This work proposes PSMOA, a novel Policy Support Multi-objective Optimization Algorithm for decentralized data replication that dynamically adapts to varying organizational requirements such as minimization or maximization of replication time, storage cost, replication based on content popularity, and load balancing while respecting policy constraints. PSMOA outperforms NSGA-II and NSGA-III in both Generational Distance (20.29 vs 148.74 and 67.74) and Inverted Generational Distance (0.78 vs 3.76 and 5.61), indicating better convergence and solution distribution. These results validate PSMOA's novelty in optimizing data replication in multi-organizational environments.

PSMOA: Policy Support Multi-Objective Optimization Algorithm for Decentralized Data Replication

TL;DR

PSMOA addresses the challenge of policy-aware, multi-objective data replication in decentralized, multi-organizational environments by integrating NSGA-III with a dynamic policy framework. It uses Entropy Weighted TOPSIS to assign objective weights and generate reference points, while a policy expression module adapts weights in response to system conditions and user constraints. The approach optimizes four objectives—replication time, cost, popularity, and load balance—under storage and bandwidth constraints, leveraging NSGA-III to maintain solution diversity and convergence. Experimental results show PSMOA achieving superior generational distance and inverted generational distance compared to NSGA-II/III, along with notable load-balancing gains and effective dynamic policy adaptation, indicating strong practical potential for federated storage systems like the WLCG.

Abstract

Efficient data replication in decentralized storage systems must account for diverse policies, especially in multi-organizational, data-intensive environments. This work proposes PSMOA, a novel Policy Support Multi-objective Optimization Algorithm for decentralized data replication that dynamically adapts to varying organizational requirements such as minimization or maximization of replication time, storage cost, replication based on content popularity, and load balancing while respecting policy constraints. PSMOA outperforms NSGA-II and NSGA-III in both Generational Distance (20.29 vs 148.74 and 67.74) and Inverted Generational Distance (0.78 vs 3.76 and 5.61), indicating better convergence and solution distribution. These results validate PSMOA's novelty in optimizing data replication in multi-organizational environments.

Paper Structure

This paper contains 30 sections, 5 equations, 4 figures, 1 table, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related work
  3. PSMOA
  4. Entropy Weighted TOPSIS
  5. Objective Weight Assignment
  6. Reference Point Generation
  7. Determine the positive ideal solution $Z^+$ and negative ideal solution $Z^-$ in the objective space
  8. Calculate the closeness of each solution to the positive and negative ideal solutions
  9. Compute the relative closeness coefficient $C_i$ for each solution
  10. Generate reference points based on the relative closeness coefficients and entropy weights
  11. Policy Expression
  12. Dynamic Policy Adaptation
  13. User-Imposed Constraints
  14. NSGA-III Algorithm for Node Selection
  15. Optimization Objectives
  16. ...and 15 more sections

Figures (4)

  • Figure 1: Decentralized data replication system model.
  • Figure 2: Performance comparison of NSGA-II, NSGA-III, and PSMOA at population size 100.
  • Figure 3: Pareto front comparisons across different system scales showing trade-offs between different objectives.
  • Figure 4: Weight changes and performance over 24-hours