Developing Analytical Framework to Measure Robustness of Peer-to-Peer Networks Niloy Ganguly.

Slides:



Advertisements
Similar presentations
Mobile Communication Networks Vahid Mirjalili Department of Mechanical Engineering Department of Biochemistry & Molecular Biology.
Advertisements

Scale Free Networks.
1 Greedy Forwarding in Dynamic Scale-Free Networks Embedded in Hyperbolic Metric Spaces Dmitri Krioukov CAIDA/UCSD Joint work with F. Papadopoulos, M.
Clayton Sullivan PEER-TO-PEER NETWORKS. INTRODUCTION What is a Peer-To-Peer Network A Peer Application Overlay Network Network Architecture and System.
Predicting Tor Path Compromise by Exit Port IEEE WIDA 2009December 16, 2009 Kevin Bauer, Dirk Grunwald, and Douglas Sicker University of Colorado Client.
How do the superpeer networks emerge? Niloy Ganguly, Bivas Mitra Department of Computer Science & Engineering Indian Institute of Technology, Kharagpur,
School of Information University of Michigan Network resilience Lecture 20.
VL Netzwerke, WS 2007/08 Edda Klipp 1 Max Planck Institute Molecular Genetics Humboldt University Berlin Theoretical Biophysics Networks in Metabolism.
Farnoush Banaei-Kashani and Cyrus Shahabi Criticality-based Analysis and Design of Unstructured P2P Networks as “ Complex Systems ” Mohammad Al-Rifai.
Hierarchy in networks Peter Náther, Mária Markošová, Boris Rudolf Vyjde : Physica A, dec
Networks. Graphs (undirected, unweighted) has a set of vertices V has a set of undirected, unweighted edges E graph G = (V, E), where.
Small-world Overlay P2P Network
ZIGZAG A Peer-to-Peer Architecture for Media Streaming By Duc A. Tran, Kien A. Hua and Tai T. Do Appear on “Journal On Selected Areas in Communications,
Peer-to-Peer and Grid Computing Exercise Session 3 (TUD Student Use Only) ‏
1 Denial-of-Service Resilience in P2P File Sharing Systems Dan Dumitriu (EPFL) Ed Knightly (Rice) Aleksandar Kuzmanovic (Northwestern) Ion Stoica (Berkeley)
Dynamic Hypercube Topology Stefan Schmid URAW 2005 Upper Rhine Algorithms Workshop University of Tübingen, Germany.
Building Low-Diameter P2P Networks Eli Upfal Department of Computer Science Brown University Joint work with Gopal Pandurangan and Prabhakar Raghavan.
presented by Hasan SÖZER1 Scalable P2P Search Daniel A. Menascé George Mason University.
1 Client-Server versus P2P  Client-server Computing  Purpose, definition, characteristics  Relationship to the GRID  Research issues  P2P Computing.
TELCOM2125: Network Science and Analysis
©Silberschatz, Korth and Sudarshan18.1Database System Concepts Centralized Systems Run on a single computer system and do not interact with other computer.
COVERTNESS CENTRALITY IN NETWORKS Michael Ovelgönne UMIACS University of Maryland 1 Chanhyun Kang, Anshul Sawant Computer Science Dept.
Peer-to-Peer and Social Networks Random Graphs. Random graphs E RDÖS -R ENYI MODEL One of several models … Presents a theory of how social webs are formed.
Freenet. Anonymity  Napster, Gnutella, Kazaa do not provide anonymity  Users know who they are downloading from  Others know who sent a query  Freenet.
Large-scale organization of metabolic networks Jeong et al. CS 466 Saurabh Sinha.
The Erdös-Rényi models
Analyzing the Vulnerability of Superpeer Networks Against Churn and Attack Niloy Ganguly Department of Computer Science & Engineering Indian Institute.
Epidemic spreading in complex networks: from populations to the Internet Maziar Nekovee, BT Research Y. Moreno, A. Paceco (U. Zaragoza) A. Vespignani (LPT-
(Social) Networks Analysis III Prof. Dr. Daning Hu Department of Informatics University of Zurich Oct 16th, 2012.
1 Reading Report 4 Yin Chen 26 Feb 2004 Reference: Peer-to-Peer Architecture Case Study: Gnutella Network, Matei Ruoeanu, In Int. Conf. on Peer-to-Peer.
Author: M.E.J. Newman Presenter: Guoliang Liu Date:5/4/2012.
P.1Service Control Technologies for Peer-to-peer Traffic in Next Generation Networks Part2: An Approach of Passive Peer based Caching to Mitigate P2P Inter-domain.
Niloy Ganguly Department of Computer Science & Engineering Indian Institute of Technology, Kharagpur Kharagpur Stability analysis of peer to peer.
Network Aware Resource Allocation in Distributed Clouds.
Percolation in self-similar networks Dmitri Krioukov CAIDA/UCSD M. Á. Serrano, M. Boguñá UNT, March 2011.
Scalable Web Server on Heterogeneous Cluster CHEN Ge.
A Routing Underlay for Overlay Networks Akihiro Nakao Larry Peterson Andy Bavier SIGCOMM’03 Reviewer: Jing lu.
Super-peer Network. Motivation: Search in P2P Centralised (Napster) Flooding (Gnutella)  Essentially a breadth-first search using TTLs Distributed Hash.
Analyzing the Vulnerability of Superpeer Networks Against Attack B. Mitra (Dept. of CSE, IIT Kharagpur, India), F. Peruani(ZIH, Technical University of.
Robustness of complex networks with the local protection strategy against cascading failures Jianwei Wang Adviser: Frank,Yeong-Sung Lin Present by Wayne.
1 Peer-to-Peer Technologies Seminar by: Kunal Goswami (05IT6006) School of Information Technology Guided by: Prof. C.R.Mandal, School of Information Technology.
Analyzing the Vulnerability of Superpeer Networks Against Attack Niloy Ganguly Department of Computer Science & Engineering Indian Institute of Technology,
Mitigation strategies on scale-free networks against cascading failures Jianwei Wang Adviser: Frank,Yeong-Sung Lin Present by Chris Chang.
Lecture 10: Network models CS 765: Complex Networks Slides are modified from Networks: Theory and Application by Lada Adamic.
P2P Group Meeting (ICS/FORTH) Monday, 28 March, 2005 A Scalable Content-Addressable Network Sylvia Ratnasamy, Paul Francis, Mark Handley, Richard Karp,
LightFlood: An Efficient Flooding Scheme for File Search in Unstructured P2P Systems Song Jiang, Lei Guo, and Xiaodong Zhang College of William and Mary.
Percolation and diffusion in network models Shai Carmi, Department of Physics, Bar-Ilan University Networks Percolation Diffusion Background picture: The.
ADVANCED COMPUTER NETWORKS Peer-Peer (P2P) Networks 1.
Percolation in self-similar networks PRL 106:048701, 2011
Data Communications and Networks Chapter 9 – Distributed Systems ICT-BVF8.1- Data Communications and Network Trainer: Dr. Abbes Sebihi.
Brief Announcement : Measuring Robustness of Superpeer Topologies Niloy Ganguly Department of Computer Science & Engineering Indian Institute of Technology,
1 NETWORKING 2012 Parallel and Distributed Systems Group, Delft University of Technology, the Netherlands May 22, 2012 Reducing the History in Decentralized.
Community structure in graphs Santo Fortunato. More links “inside” than “outside” Graphs are “sparse” “Communities”
P2P Search COP P2P Search Techniques Centralized P2P systems  e.g. Napster, Decentralized & unstructured P2P systems  e.g. Gnutella.
Network Partition –Finding modules of the network. Graph Clustering –Partition graphs according to the connectivity. –Nodes within a cluster is highly.
1 Roie Melamed, Technion AT&T Labs Araneola: A Scalable Reliable Multicast System for Dynamic Wide Area Environments Roie Melamed, Idit Keidar Technion.
Response network emerging from simple perturbation Seung-Woo Son Complex System and Statistical Physics Lab., Dept. Physics, KAIST, Daejeon , Korea.
Complex network of the brain II Attack tolerance vs. Lethality Jaeseung Jeong, Ph.D. Department of Bio and Brain Engineering, KAIST.
Dynamic Network Analysis Case study of PageRank-based Rewiring Narjès Bellamine-BenSaoud Galen Wilkerson 2 nd Second Annual French Complex Systems Summer.
Complex network of the brain II Hubs and Rich clubs in the brain Jaeseung Jeong, Ph.D. Department of Bio and Brain Engineering, KAIST.
Random Walk for Similarity Testing in Complex Networks
Data Center Network Architectures
Peer-to-Peer and Social Networks
Determining the Peer Resource Contributions in a P2P Contract
Resolution Limit in Community Detection
Peer-to-Peer and Social Networks Fall 2017
Research Scopes in Complex Network
Robustness or Network Resilience
Peer-to-Peer and Social Networks
Deterministic and Semantically Organized Network Topology
Presentation transcript:

Developing Analytical Framework to Measure Robustness of Peer-to-Peer Networks Niloy Ganguly

Client/Server Architecture Well known, powerful, reliable server is a data source Clients request data from server Very successful model WWW (HTTP), FTP, Web services, etc. Server Client Internet

Client/Server Limitations Scalability is hard to achieve (load balancing) Presents a single point of failure (fault tolerance) Network bandwidth adds the bottleneck problem Requires administration P2P systems try to address these limitations

Peer to Peer Networks All nodes are both clients and servers i.e. Servent (SERVer+cliENT) Provide and consume data Any node can initiate a connection No centralized data source “The ultimate form of democracy on the Internet” Node Internet

Peer to Peer Networks Popular medium for file sharing and other applications like IP telephony, distributed storage, publisher subscriber system,etc. Overlay Networks  Logical network above the physical p2p network  Importance of the topology of overlay networks  Spread of information in the network  Stability of the network due to dynamic nature of the peers.

Overlay Network An overlay network is built on top of physical network. Nodes in the overlay can be thought of as being connected by virtual or logical links, each of which corresponds to a path, perhaps through many physical links, in the underlying network. Examples: P2P overlay network run on top of the Internet.

Overlay Network An overlay network is built on top of physical network. Nodes in the overlay can be thought of as being connected by virtual or logical links, each of which corresponds to a path, perhaps through many physical links, in the underlying network. Examples: P2P overlay network run on top of the Internet. Overlay edge

Overlay Network An overlay network is built on top of physical network. Nodes in the overlay can be thought of as being connected by virtual or logical links, each of which corresponds to a path, perhaps through many physical links, in the underlying network. Examples: P2P overlay network run on top of the Internet. Overlay edge

Motivation Peers in the p2p system join and leave network randomly without any central coordination. Makes overlay structures highly dynamic in nature. Frequently it partitions the network into smaller fragments Communication between peers become impossible. In this work, our primary goal is to develop an analytical framework to examine the stability of the various overlay structures against dynamic movement of peers.

Overview Various Overlay Structures – Topology Dynamic movements of peers Stability criterion Analytical framework

Topology of the Overlay Networks Topology of the overlay networks can be modeled From experimentally collected data of overlay topology By various random graphs characterized by degree distribution Examples: E-R graph N number of vertices are connected with probability p. Probability of any randomly chosen node having degree k becomes Degree distribution follows Poisson distribution. Maintains homogeneous connectivity Scale free network Inhomogeneous connectivity Majority of nodes have only a few links and very few highly connected nodes control the connectivity of entire network. Degree distribution follows power law distribution

Topology of the Overlay networks Superpeer networks Small fraction of nodes are superpeers and rest are peers Each superpeer node is connected with a set of peers Superpeers are connected among themselves KaZaA adopted this kind of topology Follows Bimodal degree distribution Mathematically if otherwise Superpeer Node Peer node

Percolation and Peer Movement Movement of the peers can be modeled by various kinds of node failures in the random graph  Degree independent node failure  Probability of removal of a node is constant & degree independent  Targeted Attack  Nodes having highest connectivity is removed first  Degree dependent node failure  Probability of removal of a node is inversely proportional to the degree of that node  Peers having lower connectivity are less stable because they enter and leave network frequently.

Stability Metric - Percolation Threshold Percolation threshold is the critical fraction of nodes whose removal disintegrates the giant component into smaller fragmented components We use percolation threshold as the stability metric for our analysis. Initially all the nodes in the network are connected. Forms a single giant component.

Percolation Threshold Initial single connected component f fraction of nodes removed Giant component still exists

Percolation Threshold Initial single connected component f fraction of nodes removed Giant component still exists f c fraction of nodes removed The entire graph breaks into smaller fragments Therefore f c becomes percolation threshold

Stability Analysis We use generating function formalism to perform stability analysis. According to this formalism, general formula for the stability of the giant component with respect to any type of graph (p k ) and any kind of failure (q k ) becomes

Generating function formalism  Generating function:  Formal power series whose coefficients encode information. Here encode information about a sequence Used to understand different properties of the graph generates the probability distribution of the vertex degrees. Average degree

Stability Analysis and specifies the network topology and failure models respectively. specifies the probability of a node having degree to be present in the network after the process of removal of some portion of nodes is completed. becomes the corresponding generating function.

Stability Analysis Topology specified by Fraction of nodes removed according to specifies the probability of a node having degree k after the process of node removal

Stability Analysis Let F 1 (x) generates the distribution of outgoing edges of the first neighbor of a randomly chosen node. F 1 (x)=F’ 0 (x)/z H 1 (x) generates the distribution of the component sizes reached by following a random edge. H 1 (x) satisfies a self-consistency condition of the form H 1 (x)=1-F 1 (1)+xF 1 (H 1 (x)) Randomly selected node ‘A’ A Degree distribution of the first neighbor of ‘A’

Stability Analysis  Distribution for the component size to which a randomly selected node belongs to generated by H 0 (x) where H 0 (x)=1-F 0 (1)+xF 0 (H 1 (x)) = ++++ ……. Schematic representation of the sum rule for the connected component of vertices reached by following a random edge. (Self consistency condition)

Stability Analysis Average size of the component Which diverges when F 1 ’(1)=1 This equation states the critical condition for the stability of giant component  For any kind of graph ( p k )  Undergoing any kind of failure (1-q k )

Stability at various scenario Stability of generalized random graphs undergoing various failures  Degree Independent random failure : In this case q k =q=q c Using Therefore percolation threshold

Stability at various scenario Degree dependent failure: In this case In extreme case α = 1 Therefore according to our general formula Critical condition for percolation becomes Thus critical fraction of node removed becomes where which satisfies the above equation

Case Study:Superpeer Networks Recently superpeer networks have been adopted by many p2p systems like KaZaA. A small fraction of nodes are superpeers and rest are peers. Connectivity of superpeers are much more higher than the peers. It can be modeled by bimodal degree distribution.

Case Study:Superpeer Networks Degree independent failure: According to our framework, critical fraction for superpeer networks where r = fraction of peers Superpeer degree Average degree of the network

Case Study:Superpeer networks Degree independent failure Comparative study between theoretical and experimental results Theoretical Experimental

Observations Increase of the fraction of superpeers (specially above 15% to 20%) increases stability of the network. Experimental result indicates the optimum superpeer to peer ratio for which overlay networks becomes most stable for this kind of failure. Due to the contradiction of theoretical and practical concept of giant component, there is a little difference between theoretical and experimental results.

Case Study:Superpeer Networks Degree dependent failure: In this case, the value of which percolates the network can be derived from our general formula and becomes where Superpeer degree Average degree of the network

Case Study:Superpeer networks Degree dependent failure Comparative study between theoretical and experimental results Theoretical Experimental

Observations With the increase of superpeer degree, the value of γ c that percolates the network decreases. Thus it improves the stability of the network and the improvement follows hyperbolic trajectories. Result supports our intuitive notion of giant component.

Conclusion Contribution of our work Development of general framework to analyze the stability of p2p overlay networks. Modeling the behavior of the peers using degree independent as well as degree dependent node failure. Case Study : stability analysis of the superpeer networks. Perform a comparative study between theoretical and experimental results to show the effectiveness of our theoretical model.

Future Work  We have to perform a detailed comparative study of stability of various overlay structures.  Example: E-R networks,scale free networks, various kinds of superpeer networks like Mixed Poisson and bimodal structures.  Peer movements can be modeled by various kinds of node failures and attacks where nodes having more importance are been targeted.  Importance of a node is determined by degree centrality, betweenness centrality, eigenvector centrality etc  Finally a comparative stability analysis of all these topologies with respect to combination of different attacks and failures.

Stability Criterion Giant Component Most of the nodes in the network are connected to form a large connected component After removing a fraction of nodes from the network A large fraction of nodes still remains connected. Although average distance increases. A fraction of nodes removed from the network Percolation process Giant component

Percolation Process: Degree independent failure Occupied Node Unoccupied Node Nodes to be removed are selected at random (do not dependent on their degree)

Percolation Process: Degree independent failure After random removal of nodes, network disintegrated into disconnected components

Percolation Process: Targeted Attack Highly connected nodes Occupied Node Unoccupied Node Highly connected nodes are attacked first

Percolation Process: Targeted Attack After targeted attack, network is disintegrated into disconnected components

Percolation Process: Degree dependent failure Occupied Node Unoccupied Node Nodes to be removed are inversely proportional to its degree

Percolation Process: Degree dependent failure After degree dependent failure, network is disintegrated into spited components

Movement of peers Topology of the overlay networks can be modeled by various real world networks. Movement of peers can be represented by percolation processes in those graphs. We study the stability of various topologies by measuring the effect of percolation on the connectivity of the graph.