DECENTRALIZED SCALE-FREE NETWORK CONSTRUCTION AND LOAD BALANCING IN MASSIVE MULTIUSER VIRTUAL ENVIRONMENTS



Similar documents
A Novel Switch Mechanism for Load Balancing in Public Cloud

A discussion of Statistical Mechanics of Complex Networks P. Part I

USING SPECTRAL RADIUS RATIO FOR NODE DEGREE TO ANALYZE THE EVOLUTION OF SCALE- FREE NETWORKS AND SMALL-WORLD NETWORKS

Chapter 29 Scale-Free Network Topologies with Clustering Similar to Online Social Networks

Graph Theory and Complex Networks: An Introduction. Chapter 08: Computer networks

Big Data Analytics of Multi-Relationship Online Social Network Based on Multi-Subnet Composited Complex Network

Dmitri Krioukov CAIDA/UCSD

International journal of Engineering Research-Online A Peer Reviewed International Journal Articles available online

Complex Network Visualization based on Voronoi Diagram and Smoothed-particle Hydrodynamics

How To Understand The Network Of A Network

Complex Networks Analysis: Clustering Methods

DATA ANALYSIS IN PUBLIC SOCIAL NETWORKS

D1.1 Service Discovery system: Load balancing mechanisms

Scalable Source Routing

Introduction to Networks and Business Intelligence

Greedy Routing on Hidden Metric Spaces as a Foundation of Scalable Routing Architectures

An Optimization Model of Load Balancing in P2P SIP Architecture

International Journal of Scientific & Engineering Research, Volume 4, Issue 11, November ISSN

GENERATING AN ASSORTATIVE NETWORK WITH A GIVEN DEGREE DISTRIBUTION

Graph Theory and Complex Networks: An Introduction. Chapter 06: Network analysis

Simulating a File-Sharing P2P Network

Graph models for the Web and the Internet. Elias Koutsoupias University of Athens and UCLA. Crete, July 2003

Bioinformatics: Network Analysis

Healthcare Analytics. Aryya Gangopadhyay UMBC

Load balancing in a heterogeneous computer system by self-organizing Kohonen network

Load Balancing in Structured P2P Systems

JBoss Data Grid Performance Study Comparing Java HotSpot to Azul Zing

Why Rumors Spread Fast in Social Networks

Load Balancing in Structured Peer to Peer Systems

Load Balancing in Structured Peer to Peer Systems

Structural constraints in complex networks

Graph Mining and Social Network Analysis

Redistribution of Load in Cloud Using Improved Distributed Load Balancing Algorithm with Security

PERFORMANCE STUDY AND SIMULATION OF AN ANYCAST PROTOCOL FOR WIRELESS MOBILE AD HOC NETWORKS

Mining Social Network Graphs

DYNAMIC LOAD BALANCING IN A DECENTRALISED DISTRIBUTED SYSTEM

Detecting spam using social networking concepts Honours Project COMP4905 Carleton University Terrence Chiu

Graphs over Time Densification Laws, Shrinking Diameters and Possible Explanations

Effects of node buffer and capacity on network traffic

Seminar. Path planning using Voronoi diagrams and B-Splines. Stefano Martina

Mapping the Gnutella Network: Macroscopic Properties of Large-Scale Peer-to-Peer Systems

Link Prediction in Social Networks

A PROXIMITY-AWARE INTEREST-CLUSTERED P2P FILE SHARING SYSTEM

Research Article A Comparison of Online Social Networks and Real-Life Social Networks: A Study of Sina Microblogging

Computing Load Aware and Long-View Load Balancing for Cluster Storage Systems

Distance Degree Sequences for Network Analysis

Traffic Engineering for Multiple Spanning Tree Protocol in Large Data Centers

Load Balancing. Load Balancing 1 / 24

Enabling Multi-pipeline Data Transfer in HDFS for Big Data Applications

Efficient Content Location Using Interest-Based Locality in Peer-to-Peer Systems

Strategies for Optimizing Public Train Transport Networks in China: Under a Viewpoint of Complex Networks

CHAPTER 8 CONCLUSION AND FUTURE ENHANCEMENTS

Content. Massively Multiplayer Online Games Previous Work. Cluster-based Approach. Evaluation Conclusions. P2P-based Infrastructure

Intelligent Content Delivery Network (CDN) The New Generation of High-Quality Network

Energy Efficient Load Balancing among Heterogeneous Nodes of Wireless Sensor Network

TÓPICOS AVANÇADOS EM REDES ADVANCED TOPICS IN NETWORKS

2. Research and Development on the Autonomic Operation. Control Infrastructure Technologies in the Cloud Computing Environment

General Network Analysis: Graph-theoretic. COMP572 Fall 2009

TÓPICOS AVANÇADOS EM REDES ADVANCED TOPICS IN NETWORKS

Cloud Enabled Emergency Navigation Using Faster-than-real-time Simulation

Open Access Research on Application of Neural Network in Computer Network Security Evaluation. Shujuan Jin *

LOAD BALANCING AND EFFICIENT CLUSTERING FOR IMPROVING NETWORK PERFORMANCE IN AD-HOC NETWORKS

Graph Theory and Complex Networks: An Introduction. Chapter 06: Network analysis. Contents. Introduction. Maarten van Steen. Version: April 28, 2014

Exploring Big Data in Social Networks

A REPORT ON ANALYSIS OF OSPF ROUTING PROTOCOL NORTH CAROLINA STATE UNIVERSITY

arxiv:physics/ v1 6 Jan 2006

SOCIAL NETWORK ANALYSIS EVALUATING THE CUSTOMER S INFLUENCE FACTOR OVER BUSINESS EVENTS

The Effect of Caches for Mobile Broadband Internet Access

Dynamic Load Balancing for Cluster-based Publish/Subscribe System

Costs and Benefits of Reputation Management Systems

ModelingandSimulationofthe OpenSourceSoftware Community

The Network Structure of Hard Combinatorial Landscapes

How To Understand The Power Of Icdn

Proposition of a new approach to adapt SIP protocol to Ad hoc Networks

A Survey on Load Balancing Techniques Using ACO Algorithm

How To Build A Clustered Storage Area Network (Csan) From Power All Networks

Quality of Service Routing Network and Performance Evaluation*

Internet (IPv4) Topology Mapping. Department of Computer Science The University of Texas at Dallas

Driving Value From Big Data

On Design Principles for Self-Organizing Network Functions. ISWCS, Barcelona, August 2014

Towards a Load Balancing in a Three-level Cloud Computing Network

Transcription:

DECENTRALIZED SCALE-FREE NETWORK CONSTRUCTION AND LOAD BALANCING IN MASSIVE MULTIUSER VIRTUAL ENVIRONMENTS Markus Esch, Eric Tobias - University of Luxembourg

MOTIVATION HyperVerse project Massive Multiuser Virtual Environment (MMVE) Global-scale Open Similar to the Web D Web

TWO TIER INFRASTRUCTURE Client overlay: Data distribution PS PS PS PS PS PS PS Public Server overlay: Reliable hosting Client management

BACKBONE OVERLAY Interconnect Public Servers Requirements: Distribute world surface Distribute load Heterogeneous set of Public Serves. Consider high client dynamics

CONCEPT Voronoi diagram Server position corresponding to capacities Adapt position to dynamic load shifts Scale-free link structure

LOAD BALANCING Assign object masses and Public Server payload Set of rules applied by Public Servers Local Rules Handle local load imbalances Global Rules Handle global load imbalances

LOCAL RULES Centering Move towards center of mass B Minimize nodes at Voronoi borders A

LOCAL RULES Unload Neighbors Move towards overburdened servers / & '()*, -()*,, '()*+, -().,!!# '()*,, -(*, '()*+, -(),, # '()+, -(), $ " '(),, -()*, % '()1, -()*, ')2)'45678)-)2)9:55)-;;8))))))))2)6<:=567:7

LOCAL RULES Keeping the massorder Having regions with high masses hosted by powerful servers 1. $%&'() *%&')) $%&'/) *%&) $! *!! $ # * # $%&')) *%&) # $%&'() *%&+) $%&/) *%&(), " $%&()) *%&'() #! - $%&?) *%&') $&2&$45678&*&2&9:55&*;;8&&&&&&&&2&96< =9>

LOCAL RULES Swapping :; Hotspot: The spot in a voronoi cell with the biggest mass within a certain radius :; := ' (?)A; & (?)<;; <; % (?)=;! (?)A; # (?)@; <= $ (?)>=; :; " (?)A= >;

GLOBAL RULES Jumping Epidemic aggregation Active Search ()*+,*-./1,2 #" $ & ' Active Pull %!" (4.'""

LOAD BALANCING Simulation Video:

OVERLOAD MASS Overload Mass Ratio 45 4 5 25 2 15 1 5 Overall Load: 8% 1 2 4 5 6 7 Step Overload Mass Ratio 4 5 25 2 15 1 5 Overall Load: 6% 1 2 4 5 6 7 Step Overload Mass Ratio 25 2 15 1 5 Overall Load: 4% 1 2 4 5 6 7 Step Self-organization scheme Unified distribution

OVERLOAD DURATION 45 4 5 Duration in Steps 25 2 15 1 5 1 2 4 5 6 7 Step Unified scheme, 4% Unified scheme, 6% Unified scheme, 8% Self-organization scheme, 4% Self-organization scheme, 6% Self-organization scheme, 8%

HOTSPOT ACCURACY 1 9 8 Accuracy 7 6 5 4 1 2 4 5 6 7 Step 4% 6% 8%

SCALE-FREE LINK STRUCTURE Immediate Neighbors & Fare Neighbors Scale-free Node Degree Distribution: P(k) k -γ (γ (2,)) Expect scale-free capacity distribution Observed in WWW Π(p i )= p i p max probability distribution a joining node estab-

SCALE-FREE LINK STRUCTURE Immediate Neighbors & Fare Neighbors Scale-free Node Degree Distribution: P(k) k -γ (γ (2,)) Expect scale-free capacity distribution Observed in WWW Π(p i )= p i p max probability distribution a joining node estab-

SCALE-FREE LINK STRUCTURE Immediate Neighbors & Fare Neighbors Scale-free Node Degree Distribution: P(k) k -γ (γ (2,)) Assume scale-free capacity distribution Observed in WWW Π(p i )= p i p max probability distribution a joining node estab-

ALGORITHM N

ALGORITHM N

N Web sites increases faster due to their high pro time, the providers take care for allocating capacities. That means, the scale-free link s existence of hubs with sufficient capacities em organizing manner from the different popular Transferring this observation to our scenario o environment, it means, that Public Servers ho lar world objects automatically exhibit higher this is automatically ensured by the object the scale-free capacity distribution thus emerg we just have to construct the link structure this reason, we establish links to Public Se capacities, with a higher probability. Hence connected to an existing node i with probabil on the payload p i of i and the maximum network p max : ALGORITHM Π(p i )= p i p max Based on this probability distribution a joi lishes M (M >; M is a fixed parameter o

γ = 2.1; M = 1 1 1 γ = ; M = 1 1 γ = 2.7; M = 1 1 1 1 Nodes 1 Nodes 1 Nodes 1 1 1 1 1 1 1 1 1 1 Degree 1 1 1 1 1 1 Degree 1 1 1 1 1 1 Degree γ = 2.1; M = 2 1 1 γ = ; M = 2 1 γ = 2.7; M = 2 1 1 1 NODE DEGREE Nodes 1 Nodes 1 Nodes 1 1 1 1 DISTRIBUTION 1 1 1 1 1 1 Degree 1 1 1 1 1 1 Degree 1 1 1 1 1 1 Degree 1 γ = 2.1; M = 1 γ = ; M = 1 γ = 2.7; M = 1 1 1 Nodes 1 Nodes 1 Nodes 1 1 1 1 1 1 1 1 1 1 Degree 1 1 1 1 1 1 Degree 1 1 1 1 1 1 Degree 5 Nodes 1 Nodes

POWER LAW EXPONENT.2 M = 1.2 M = 2.2 M = Power-Law Exponent 2 Power-Law Exponent 2 Power-Law Exponent 2 1 4 7 1 1 4 7 1 1 4 7 1 γ = 2.1 γ = γ = 2.7

CLUSTERING COEFFICIENT.4 γ = 2.1; M = 1.4 γ = ; M = 1.4 γ = 2.7; M = 1 Clustering Coefficient.5..25.2.15.1.5 Clustering Coefficient.5..25.2.15.1.5 Clustering Coefficient.5..25.2.15.1.5 1 4 7 1.4 γ = 2.1; M = 2.4 γ = ; M = 2.4 γ = 2.7; M = 2 Clustering Coefficient.5..25.2.15.1.5 Clustering Coefficient.5..25.2.15.1.5 Clustering Coefficient.5..25.2.15.1.5 1 4 7 1.4 γ = 2.1; M =.4 γ = ; M =.4 γ = 2.7; M = Clustering Coefficient.5..25.2.15.1.5 Clustering Coefficient.5..25.2.15.1.5 Clustering Coefficient.5..25.2.15.1.5 Scale-Free Graph Random Graph

AVERAGE SHORTEST PATH Average Shortest Path γ = 2.1; M = 1.4.2 2 Average Shortest Path γ = ; M = 1.4.2 2 Average Shortest Path γ = 2.7; M = 1.4.2 2 1 4 7 1 Average Shortest Path γ = 2.1; M = 2.4.2 2 Average Shortest Path γ = ; M = 2.4.2 2 Average Shortest Path γ = 2.7; M = 2.4.2 2 1 4 7 1 Average Shortest Path γ = 2.1; M =.4.2 2 Average Shortest Path γ = ; M =.4.2 2 Average Shortest Path γ = 2.7; M =.4.2 2 Scale-Free Graph Random Graph

NETWORK DIAMETER 4.5 γ = 2.1; M = 1 4.5 γ = ; M = 1 4.5 γ = 2.7; M = 1 4 4 4 Diameter.5 Diameter.5 Diameter.5 2.5 2.5 2.5 1 4 7 1 4.5 γ = 2.1; M = 2 4.5 γ = ; M = 2 4.5 γ = 2.7; M = 2 4 4 4 Diameter.5 Diameter.5 Diameter.5 2.5 2.5 2.5 1 4 7 1 4.5 γ = 2.1; M = 4.5 γ = ; M = 4.5 γ = 2.7; M = 4 4 4 Diameter.5 Diameter.5 Diameter.5 2.5 2.5 2.5

CHURN SIMULATIONS = 1.2 M = 1 Power Law Exponent. 2. γ = 2.1 γ = γ = 2.7 1 2 4 5 Step

CHURN SIMULATIONS.4 γ = 2.1; M = 1.4 γ = ; M = 1.4 γ = 2.7; M = 1 Clustering Coefficient.5..25.2.15.1.5 Clustering Coefficient.5..25.2.15.1.5 Clustering Coefficient.5..25.2.15.1.5 1 2 4 5 1 2 4 5 1 2 4 5 Step Step Step Scale-Free Graph Random Graph.4 γ = 2.1; M = 1.4 γ = ; M = 1.4 γ = 2.7; M = 1 Average Shortest Path.2 2 1 2 4 5 Average Shortest Path.2 2 1 2 4 5 Average Shortest Path.2 2 1 2 4 5 Step Step Step Scale-Free Graph Random Graph 4.5 γ = 2.1; M = 1 4.5 γ = ; M = 1 4.5 γ = 2.7; M = 1 4 4 4 Diameter.5 Diameter.5 Diameter.5 2.5 2.5 2.5 1 2 4 5 1 2 4 5 1 2 4 5 Step Step Step

CONCLUSION Backbone overlay for MMVE scenario Dynamic load balancing Scale-free small-world network Simulation results show viability for intended scenario

THANK YOU FOR YOUR ATTENTION...... QUESTIONS?