Availability Aware Continuous Replica Placement Problem

TL;DR

This paper introduces AACRPP, a new approach for updating replica placement in distributed systems considering current state, user requests, and node failures to improve availability and performance.

Contribution

It proposes an availability-aware extension to the continuous replica placement problem using modified greedy heuristics and simulation for evaluation.

Findings

AACRPP effectively adapts replica placement considering failures and requests.

Simulation results show improved availability and load balancing.

The method outperforms traditional static and dynamic replica placement approaches.

Abstract

Replica placement (RP) intended at producing a set of duplicated data items across the nodes of a distributed system in order to optimize fault tolerance, availability, system performance load balancing. Typically, RP formulations employ dynamic methods to change the replica placement in the system potentially upon user request profile. Continuous Replica Placement Problem (CRPP) is an extension of replica placement problem that takes into consideration the current replication state of the distributed system along with user request profile to define a new replication scheme, subject to optimization criteria and constraints. This paper proposes an alternative technique, named Availability Aware Continuous Replica Placement Problem (AACRPP).AACRPP can be defined as: Given an already defined replica placement scheme, a user request profile, and a node failure profile define a new replication scheme, subject to optimization criteria and constraints. In this effort we use modified greedy heuristics from the CRPP and investigated the proposed mechanism using a trace driven java based simulation.