Scribd
Upload
13 views

CSC 412 (Net-Centric Computing)

AI

Uploaded by

Tobi O
Copyright
© © All Rights Reserved
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
13 views

CSC 412 (Net-Centric Computing)

AI

Uploaded by

Tobi O
Copyright
© © All Rights Reserved
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 6

DESIGN AND IMPLEMENTATION OF LOAD

BALANCING ALGORITHMS FOR


DISTRIBUTED SYSTEMS

Name: OGUNBAMERU ANTHONY OLUWATOBI


Matric Number: CSC/2021/01/024
Date: 25TH November, 2024
ABSTRACT

This research focuses on improving load balancing in distributed and


grid systems. It addresses challenges such as dynamic workloads,
resource heterogeneity, and latency by integrating traditional and
modern computational approaches. Two novel algorithms are
introduced: a decentralized, sender-initiated algorithm for grid systems
and an adaptive algorithm for distributed systems. Simulations of these
algorithms demonstrate improved system response times, reduced
latency, and better resource utilization.

TABLE OF CONTENTS
TITLE PAGE
Introduction 3
Objectives 3
Literature Review 4
Methodology 4
Results 5
Summary 5
Conclusion 6
Recommendations 6
References 6
INTRODUCTION

Background:
Distributed and grid systems rely on resource sharing among nodes, but their goals
differ. Distributed systems focus on balancing workloads across processors, while
grid systems extend this functionality to geographically dispersed resources. Load
balancing is crucial to achieve optimal performance.

Problem Statement:
Existing load-balancing methods struggle in dynamic and heterogeneous
environments. Traditional approaches often fail in grid systems due to workload
variability and differing node capacities, while distributed systems face challenges
in adapting to rapidly changing conditions.

Scope and Significance:


This research aims to design algorithms that enhance resource utilization, reduce
latency, and ensure reliability in both distributed and grid systems.

OBJECTIVES
 To develop efficient load-balancing algorithms for distributed and grid
systems.
 Analyze existing load-balancing methods and identify their limitations.
 Develop and implement two novel algorithms:
 An adaptive algorithm for distributed systems.
 A sender-initiated algorithm for grid systems.
 Evaluate these algorithms under diverse system conditions.
LITERATURE REVIEW

Load Balancing in Distributed Systems:


Adaptive algorithms offer flexibility in responding to changes. Existing
approaches include static, dynamic, and hybrid models, each with
unique advantages and challenges.
Taxonomy of Load Balancing Algorithms:
A comprehensive review of algorithm components (e.g., information and
transfer policies) and attributes (e.g., scalability and autonomy)
highlights the need for adaptable and dynamic algorithms.

METHODOLOGY AND ALGORITHM DESIGN

Algorithm 1: Adaptive Load Balancing for Distributed Systems

 Information Policy: Nodes exchange dynamic load metrics.

 Transfer Policy: Tasks migrate based on system state.

 Adaptability: Parameters are dynamically tuned for optimal performance.

 Simulation Models: Tested on homogeneous/heterogeneous and


disk-based/diskless configurations.
RESULTS

Performance of Adaptive Algorithm in Distributed Systems:


 Significant performance gains in heterogeneous environments.
 Effectively handled high workload variability.
 Challenges included balancing communication overhead with performance
improvements.
 Performance of Sender-Initiated Algorithm in Grid Systems:
 Achieved faster response times and lower latency compared to traditional
methods.
 The decentralized approach enhanced scalability but occasionally delayed
task migration.

Comparison of Algorithms:
The sender-initiated algorithm excelled in grid systems, effectively managing
heterogeneity. Conversely, the adaptive algorithm proved robust in distributed
systems with dynamic workloads.

SUMMARY

This research presents two algorithms tailored for load balancing in distributed and
grid systems. The adaptive algorithm efficiently handles dynamic workloads in
distributed systems, while the sender-initiated algorithm optimizes performance in
grid environments.
CONCLUSION

Both algorithms showed significant improvements in performance, adaptability,


and scalability, contributing valuable solutions to computational resource
management.

RECOMMENDATIONS

Future work should focus on real-world implementation, middleware-independent


modules, and hybrid algorithm development.

REFERENCES
1. K. Benmohammed-Mahieddine, "An Evaluation of Load Balancing
Algorithms for Distributed Systems," PhD Thesis, University of Leeds,
1991.
2. S. S. Patil, P. Singh, "Efficient Load Balancing Algorithm in Grid
Environment," Int’l J. of Computer Science and Information Tech., vol. 2,
no. 5, pp. 2159–2164, 2011.

You might also like