In this thesis existing mathematical models for P2P systems are presented and evaluated. Thereby the search efficiency in structured and unstructured P2P overlays, the features and restrictions in P2P streaming applications, the service capacity in P2P file sharing systems, content download and replication times in P2P networks and many other issues are investigated.
Furthermore the new extended model, possibly combining the most essential characteristics of P2P systems in a consistent way is presented. The characteristics
observed in the new model are categorized in three groups according to the P2P system properties being described: the overlay parameters, the characteristics of participating peers and the resource and service characteristics.
Thereafter the objectives, regulated by P2P applications’ users and providers, and aspects of underlying P2P systems they are interested in are observed. We consider the P2P applications Skype, Joost and KaZaA, differentiating between the application users and providers, and analyzing the technical view on the P2P system characteristics.
A wide variety of existing P2P systems integrating the insights of distributed systems, databases, complexity theory and many other research areas raised inconsistencies and incompatibilities in used terminologies and abstractions. Therefore providing P2P systems interoperability and creating a common model applicable for all P2P systems became a desirable goal. In this thesis the existing P2P layer architectures considering P2P systems as a set of layers ordered according to their increasing degree of abstraction are presented. Furthermore each model layer, its input, output and functionality are described individually.
Table of Contents
1 Introduction and Motivation for the Thesis
1.1 Field of Research
1.2 Contribution and Outline of this Thesis
1.3 Model Definition and Classification
2 Related Work for Existing Mathematical P2P Models
2.1 Relating Query Popularity and File Replication in Gnutella
2.2 Random Walk Search Algorithms in P2P Networks
2.3 Stochastic Fluid Theory for P2P Streaming Systems
2.4 Service Capacity of P2P Networks
2.5 An Analytic Framework for Modeling P2P Networks
2.6 Special Issue on P2P Networking and P2P Services
2.7 BubbleStorm
2.8 Controlling the Cost of Reliability in P2P Overlays
2.9 A Reference Architecture for Overlay Networks
2.10 Efficient Resource Virtualization and Sharing Strategies
2.11 Summary
3 The New Mathematical Model of P2P Systems
3.1 Overlay Parameters
3.1.1 The Basic Notations
3.1.2 Overlay Functionality
3.1.3 Routing in P2P Overlays
3.1.4 The Random Walk Search Algorithm
3.1.5 The Service Capacity
3.1.6 The Identifier Space
3.1.7 The Super Peer Network Design
3.2 The Characteristics of Participating Peers
3.2.1 The Basic Notations
3.2.2 The Delay Experienced by Peers
3.2.3 The Request Processing
3.2.4 The Peers’ Heterogeneity
3.3 The Resource and Service Characteristics
4 Extending the Mathematical Model to the Application Layer
4.1 P2P Communication: Skype
4.1.1 Skype Users
4.1.2 The Overlay Operators
4.1.3 The Participating Peers
4.2 P2P Television: Joost
4.2.1 The Joost Users
4.2.2 The Joost Network Operators
4.3 P2P File Sharing: KaZaA
4.3.1 KaZaA Users
4.3.2 KaZaA Providers
4.3.3 Participating Peers
4.4 Internet Service Providers
4.4.1 Peering
4.4.2 Communication Quality
5 P2P System Layers
5.1 Introduction
5.2 P2P Reference Architectures
5.2.1 JXTA
5.2.2 Reference Model for P2P Overlays
5.2.3 A Reference Interface for Structured P2P Systems
5.3 User Level
5.4 Application Layer
5.5 P2P Service Layer
5.6 Key-Based Routing
5.7 Overlay Network
5.8 Network Layer
6 Case Study: Modeling the GLOBASE.KOM Overlay
6.1 Introduction
6.2 Basic Notations
6.3 The Overlay Tree Structure
6.3.1 The Simplified Model of a Complete Tree
6.3.2 Dependencies of Tree Structure on the Load Parameters: Implementation Issues
6.3.3 Evaluation Results
6.4 Interconnections and Their Effect on Overlay’s Performance
6.4.1 Super Peer Failures
6.5 GLOBASE.KOM Operations
7 Conclusion
7.1 Summary and Results of This Thesis
7.2 Future Work
7.3 Contribution in the Related Research Area
Research Objectives and Core Themes
This thesis aims to bridge the gap in existing literature by providing a comprehensive overview and structured classification of mathematical models for P2P systems, ultimately developing an extensive new model that captures essential characteristics across multiple functional layers, including a specific case study on the GLOBASE.KOM overlay.
- Systematic review and evaluation of existing mathematical models for P2P networks.
- Identification and integration of essential P2P characteristics into a unified mathematical framework.
- Extension of mathematical modeling to application-layer scenarios such as Skype, Joost, and KaZaA.
- Formal definition of a six-layer functional architecture for P2P systems.
- Validation of the GLOBASE.KOM overlay model through analytical and simulation comparisons.
Excerpt from the Book
1.1 Field of Research
Peer-to-peer (P2P) computing became increasingly important in the last years, providing new exciting opportunities for business, communication and entertainment. Deployed in the financial, content sharing, business-to-business, governmental and many other markets, P2P applications cause up to 80% of the traffic in the internet. Therefore improving the performance of P2P systems has become a desirable goal and a hot research topic.
In order to evaluate the results of P2P research various approaches like simulations, testing or mathematical models can be deployed. Due to the large amount of peers participating in P2P systems, simulation as the most scalable evaluation approach is often used. However simulations require a high time overhead while not all details can be considered. Testing enables highly detailed evaluations, but is expensive and not scalable. Thus mathematical models present a less expensive alternative for detailed evaluations of P2P systems. Furthermore mathematical models can be applied in order to analyze the P2P system performance and system parameters affecting it.
Numerous mathematical models considering diverse performance aspects of P2P systems like content downloading times, the resource search delay and overhead, load balancing, content availability and many others already exist. Most models are focused just on few specific characteristics of P2P systems and do not consider other significant aspects. Thus a model typically covers just one of the P2P functional layers like network, application, overlay, user, key-based routing or service layers. Therefore creating a more extensive model taking more characteristics of P2P systems and more functional layers into account would facilitate the evaluation and analysis of P2P systems considerably and can be regarded as a desirable goal.
Summary of Chapters
1 Introduction and Motivation for the Thesis: This chapter establishes the relevance of P2P systems and outlines the thesis goals, including the creation of an extensive model and the specific focus on the GLOBASE.KOM overlay.
2 Related Work for Existing Mathematical P2P Models: This chapter provides a critical survey of existing mathematical models, categorizing them by performance metrics and their respective P2P functional layers.
3 The New Mathematical Model of P2P Systems: This chapter introduces a unified model that consolidates essential P2P characteristics, organized by overlay parameters, peer characteristics, and resource/service properties.
4 Extending the Mathematical Model to the Application Layer: This chapter applies the model to specific real-world applications including Skype, Joost, and KaZaA, while also analyzing the role of ISPs in P2P networks.
5 P2P System Layers: This chapter presents a standardized six-layer architecture for P2P systems, detailing the inputs, outputs, and functionality inherent to each level of abstraction.
6 Case Study: Modeling the GLOBASE.KOM Overlay: This chapter conducts a deep-dive analysis of the hierarchical GLOBASE.KOM overlay, modeling its tree structure and evaluating operations like lookup and area search.
7 Conclusion: This chapter summarizes the research achievements, confirms the validity of the developed models against simulation data, and suggests pathways for future investigation.
Keywords
P2P Systems, Mathematical Models, Overlay Networks, Service Capacity, GLOBASE.KOM, Peer Heterogeneity, Application Layer, Network Architecture, Load Balancing, Hierarchical Tree, Simulation, Performance Evaluation, Resource Virtualization, Fault Tolerance, Query Search Time.
Frequently Asked Questions
What is the primary focus of this research?
The work focuses on creating a comprehensive mathematical framework to model and analyze the performance characteristics of Peer-to-Peer (P2P) systems, bridging the gap between various specialized existing models.
Which functional areas are covered by the model?
The model covers the full stack of P2P architecture, including the user level, application layer, P2P service layer, key-based routing, overlay network, and physical network layer.
What is the main objective of the thesis?
The core objective is to move beyond models that focus on single characteristics and to build an extensive mathematical model that integrates essential properties of P2P systems for better evaluation and analysis.
What methodology does the author employ?
The research utilizes analytical mathematical modeling to derive system performance metrics, which are subsequently validated by comparing the theoretical results with computer simulations.
What specific overlay is modeled in the case study?
The case study focuses on GLOBASE.KOM, a hierarchical, tree-based P2P overlay specifically designed for efficient, location-based searches.
What distinguishes this model from previous ones?
Unlike most existing models that concentrate on very specific parameters, this work provides a holistic view by considering dependencies across different functional layers and introducing interconnections to mitigate the limitations of standard tree-based structures.
How does the model handle peer churn?
Peer churn is addressed by introducing specific notations for average arrival and departure rates, as well as incorporating failure detection mechanisms like keep-alive messages within the GLOBASE.KOM overlay model.
What role does peer heterogeneity play in this work?
Peer heterogeneity is a central theme; the model categorizes peers into classes based on their upload capacities and processing power, using this to optimize load balancing and search efficiency.
- Quote paper
- Inna Kotchourova (Author), 2007, Evaluation and Extension of Mathematical Models of P2P Systems, Munich, GRIN Verlag, https://www.grin.com/document/87906
