iLab P2P Networks Dirk Haage Chair for Network Architectures and - PowerPoint PPT Presentation

Network Architectures and Services, Georg Carle Faculty of Informatics Technische Universität München, Germany iLab² P2P Networks Dirk Haage Chair for Network Architectures and Services Department of Computer Science Technische Universität München http://www.net.in.tum.de

Motivation Lets work together and be our own network Today, P2P causes more than 50 % of all traffic on the Internet (2007) Internet  A need to provide services independent  More and more private users on from commercial or dedicated server the Internet (further away from the providers infrastructur)  Application-specific network structures  Powerful private end systems instead of machine location-based addressing  Flatrates with always-on users For instance, friends want their computers to be together.  Why not their own Why waste these resources? network - their own overlay network? ILab2 2

Term: Peer-to-Peer Peer-to-Peer systems  Distributed systems that consist of equals (peers) with no predefined distinction between client and server and no dedicated servers or central authority. Characteristics  Peer-to-Peer networks are decentralized and take advantage of resources at the edge of the Internet, say the computers of users, the users, etc.  End systems do not primarily serve the purpose of the Peer-to-Peer system.  their resources must not be exhausted by the Peer-to-Peer network  Computers are not always-on.  environment is less stable and more dynamic than in the traditional client-server case. ILab2 3

Peer-to-Peer or not Peer-to-Peer Auctions / Ebay  Peer-to-Peer  Money and goods exchange (nothing to do with the network)  Not Peer-to-Peer  The platform itself (Auctions, Accounts, Information transfer) and its Information Management Skype  Peer-to-Peer  Lookup, User Interaction, Data Exchange  Not Peer-to-Peer  Login, Account Management Many Peer-to-Peer systems are not purely Peer-to-Peer. ILab2 4

Some terms from Graph Theory v 1  Graph G=(V,E) v 2 v 3 v 4  Vertex set V = {v 1 , v 2 , …v n } v 5 v 6  We usally say nodes. Graph G Vertex set V  n = |V| e 3 e 4 e 6  Edge set E = {e 1 , e 2 , …e m } e 2 e 5 e 7 e 2 e 1  We usually say links.  Edge set E m = |E|  Can have attributes like distance, etc. ILab2 5

Some terms from Graph Theory Distance d(i,j)  2 distance v 1  d(v1,v6)=5 Shortest path between nodes v i and v j 1 4 1 1 2 2 Diameter D of G   Longest distance in graph G v 6 3 diameter(G)=5 Degree   Node degree = number of edges adjacent to node  Degree of a graph = max. node degree v 5 A graph is connected if there is a path from any node in  degree (v 5 ) = 3 the graph to any other node in the graph. A graph is k-connected if any k-1 nodes can be  removed without causing the resulting subgraph to become disconnected. ILab2 6

Peer-to-Peer network Peers Underlay Underlay  Provides connectivity between all peers in the Peer-to-Peer network (overlay). Peers V = {v 1 , v 2 , …v n }  Peers are the nodes of the graph G.  Peers may have a name (identities are usually necessary).  The set of edges E needs to be created by the Peer-to-Peer algorithms.  The graph needs to be connected.  The structure should be good for the purpose of the Peer-to-Peer system. ILab2 7

P2P network is not static – Peers join and leave Node joins   Needs to be added to the network join ?  Usually via some node in the network already known (rendezvous point, list/cache of nodes) Node leaves   Important to keep the leave ? graph connected  Better not rely on a single node that could leave anytime disconnected when this node fails or leaves. How to organize such a  network? 2-connected --  e.g. k-connected graph each node can be removed without disconnecting the graph ILab2 8

Application Requirements Peers Multicast from the white game server to its peers. Is this a good graph for fast delivery?  No, a balanced tree allows O(logn) diameter. Underlay Application Peer-to-Peer networks are usually created for an application or application  scenario.  Filesharing  File Distribution  Instant Messaging and Voice-over-IP  Multicast  Peer-to-Peer Video Streaming  Anonymous communication and services  … The application is the purpose of the Peer-to-Peer network.  The application and its requirements determine if a given graph is a good or a  bad choice. ILab2 9

Operational aspects  Find someone to  get something  use a service  interact  interact for a cooperative service or goal  maintain network  Find something (item, data, information, etc.) to  get it  set it  Interact with other nodes to cooperatively  provide a service  share resources  run an algorithm  … ILab2 10

Network Architectures and Services, Georg Carle Faculty of Informatics Technische Universität München, Germany iLab² Network Coordinate Systems Dirk Haage Chair for Network Architectures and Services Department of Computer Science Technische Universität München http://www.net.in.tum.de

Main Goal: Localization of Node ILab2 12

Main Goal: Localization of Node  Choosing of servers  Load balancing between hosting location  Choose nearest instance of a service (anycast)  Locate nearest peers in P2P networks  Content delivery networks  Online games (gameserver)  Resource placement in distributed systems  TO  Optimization of application layer multicast trees  … ILab2 13

What it isn’t  Provide location-based services  Local advertisements  Extend/reduce service for local/non-local users (e.g. IPTV often restricted to country boundaries)  Find friends, coworkers, …  Google Latitude  …  For this, you use GeoIP or similar approaches  GPS, Cellular Positioning, Triangulation, etc. ILab2 14

Network coordinates  Latencies between nodes as a metric for distance  Round trip time • Simplest measurement at all (ping) • Most accurate (only one clock involved) • Similar to real distance (propagation speed nearly constant)  How to get?  Simple approach: Measurements between all pairs of nodes  O(n²)  Does not scale (cannot be used for large networks)  Rely on actual traffic  hybrid measurement  Normally no traffic to all nodes available • Active measurements (even worse scaling)  You want to know the distance to a node without having to communicate with it in the first place ILab2 15

Network coordinates (II)  Measure the distances to some neighbors  Neighbors might be known hosts, not near hosts  Calculate a artificial coordinate in a metric space  Metric space = distance between nodes can be calculated  E.g. Euclidean n-space  Approximate the latency  Distance between nodes in the coordinate system is approximation to the latency  Abstract definition:  Embed network graph into a metric space  Metric embedding/ graph embedding ILab2 16

Example Internet Euclidean space (2D) (x1,y1) A RTT(A,D) A (x4,y4) D D d(B,A) RTT(D,B) RTT(D,C) (x2,y2) C C B B (x3,y3) RTT(B,C) Measured distance Estimated distance       d ( B , A ) ( x , y ), ( x , y ) x x , y y 2 2 1 1 1 2 1 2 ILab2 17

Network coordinates (III)  Advantages  Small overhead • Only requires small number of measurements • No additional traffic (application traffic = measurement traffic) • Piggy-back the coordinate information  Each host can calculate the distance to every other host • Only requires the coordinates  Design goals  Accuracy: small error for RTT estimations  Scalability: large-scale networks, small overhead, no bottlenecks  Flexibility: adapt coordinates to network changes  Stability: no drift, oscillation of coordinates  Robustness: small impact of error by malicious nodes, nodes with high errors ILab2 18

Triangle inequality  Intuition: direct latency between 2 nodes should be smaller than any indirection   d ( a , b ) d ( b , c ) d ( a , c )  Triangle inequality violations (TIV) inherent to Internet routing structure  Selective/ private peering  Hot potato routing  Link metric ≠ latency  Asymmetric links (e.g. DSL, UMTS)  TIVs are common  >85% of all host pairs part of a TIV  For 20-35% exists a path that is at least 20% shorter (Traces: King, Azureus) ILab2 19

Triangle inequality (II)  Possible spaces for embedding are metric  Distance function satisfies triangle inequality  Embedding can not be exact  Number and weight of TIVs limits embedding quality A A 26ms 22ms 19ms 17ms Embedding B C B C 38ms 53ms ILab2 20