Chapter 1: Introduction

1
Peer-to-Peer Networks Prashant Dewan Thursday,
16 September 2004
2
Introduction

• Client-Server Networks
• Peer-to-Peer Networks
• Hybrid Networks

3
Client-Server
4
Client-Server Networks

• Most common architecture
• Centralised model for data storage, security,
  running applications and network administration
• Example Mail Servers, Web Servers, News Servers

5
Client-Server Networks (2)

• Based on a scaleable model
• Users network servers
• Provide services such as printing, email etc
• Allow a high level of security to be implemented
• Can be centrally managed

6
Client-Server Pros Cons

• Advantages
• Networked web of computers
• Inexpensive but powerful array of processors
• Open systems
• Grows easily
• Individual client operating systems

• Disadvantages
• Maintenance nightmares
• Support tools lacking
• Retraining required

7
Peer-to-Peer Networks
Hub
Stations, No Server, Equal Priority Democratic
8
Peer-to-Peer Networks (2)

• Resources shared in a decentralised manner
• Shared resources include files and printers
• Should be used where nodes less than ten
• Files are not stored centrally, e.g. Kazaa
• Allows easy node-to-node communication

9
Peer-to-Peer Networks (3)

• Support is usually part of OS
• Sharing of files is responsibility of each
  participant
• Participants form a Workgroup (Windows)
• Workgroup is assigned a name important where
  there are multiple workgroups (eg dept.s)

10
Advantages

• Easy to configure
• No requirement for server hardware/software
• Users can mange their own resources
• No need for a network administrator
• Reduce total cost

11
Disadvantages

• Provide a limited number of connections
• May slow performance of nodes
• Do not allow central management
• Do not have a central store of files
• Users responsible for managing own resources
• Offers very poor security

12
P2P Issues

• Organize, maintain overlay network
• node arrivals
• node failures
• Resource allocation/load balancing
• Resource location
• Network proximity routing

13
System Architecture

• How to organize participating nodes?

Simple unstructured approach e.g., Gnutella
Pure P2P
Structured DHTs e.g., Pastry, Tapestry
Hybrid unstructured approach e.g., Napster
Hybrid P2P
Structured DHTs with some centralization
14
Routing Strategies

• Given a query, how to route it to its destination?

Broadcast each query is broadcast to all its
neighbors. (E.g. Gnutella)
Random walk Query is routed to one of its
neighbors randomly (directed random walk)
DHT Query is routed by looking up a routing
table. (Pastry, Tapestry, Chord)
15
Typical systemsNapster
central index
...
16
Typical Systems--Napster

• Central index server
• Hybrid approach C/S P2P
• Pros
• Simple yet efficient approach
• Bounded query lookup time
• Cons
• Scalability
• Reliability

17
Typical Systems--Gnutella
18
Typical Systems--Gnutella

• No central server, purely decentralized P2P
• Constrained broadcast mechanism (TTL)
• Pros
• Simple yet efficient
• Fault-tolerant
• Cons
• High bandwidth consumption
• Scalability

19
Typical systems--DHTs
Distributed application
data
get (key)
Put (key, data)
Distributed hash table

• DHT provides the information look up service for
  P2P applications.
• Nodes uniformly distributed across key space
• Nodes form an overlay network
• Nodes maintain list of neighbors in routing table

20
Typical Systems--Pastry

• 128 bit node identifier
• Pastry routes message to node with node id
  numerically closest to key value. Like a
  distributed hash table.
• Routing is based on id prefixes
• Nodes forward messages to a node closer to
  keyID
• Closer in terms of length of common prefix
• If current nodeID shares first n digits with
  keyID
• Forward to a node that shares gt n prefix digits
  with keyID
• If no longer common prefix nodeID is in routing
  table
• Then forward to a node that is numerically closer
  to keyID.
• N nodes, at most O(log(2b)N) messages along the
  routing path

21
Pastry

• One primitive
• route(M, X) route message M to the live node
  with nodeId closest to key X
• nodeIds and keys are from a large, sparse id
  space

22
Distributed Hash Tables (DHT)
nodes
k1,v1
k2,v2
k3,v3
P2P overlay network
Operations insert(k,v) lookup(k)
k4,v4
k5,v5
k6,v6

• p2p overlay maps keys to nodes
• completely decentralized and self-organizing
• robust, scalable

23
Pastry Leaf sets

• Each node maintains IP addresses of the nodes
  with the L/2 numerically closest larger and
  smaller nodeIds, respectively.
• routing efficiency/robustness
• fault detection (keep-alive)
• application-specific local coordination

24
Pastry Routing procedure
if (destination is within range of our leaf set)
forward to numerically closest member else let
l length of shared prefix let d value of
l-th digit in Ds address if (Rld exists)
forward to Rld else forward to a known
node that (a) shares at least as long a
prefix (b) is numerically closer than this node
25
Typical Systems--Pastry

• Example in base 4
• Each row down the table, the shared prefix is one
  digit longer. The next digit after the shared
  prefix is the column number
• nodeId 10233102
• E.g. Supposing
• 1. key10233-2-30
• 2. key102-1-2300

26
Typical Systems--Pastry

• Designed by Microsoft Research and Rice
  University
• Pros
• Scalable distributed modest-sized routing table
• Robust purely distributed P2P
• Cons
• Expensive self-organization O (log n) messages
• inadequateness for keyword searching exact file
  names needed for query

27
Hybrid Networks Advantages

• Client-server apps are still centrally managed
• Users can assign loacl access to their resources
• Workgroups can manage resources without need for
  assistance from network administrator

28
Hybrid Networks

• Incorporates the best features of workgroups in
  Peer-to-Peer with the performance, reliability
  and security of server-based systems.
• Allows access to central resources, but also
  allows users to function at the Peer-to-Peer
  level
• Users do not have to log in to a central server

29
P2P Security
30
Failure vs. Attack

• Failure
• Random failure of nodes and/or infrastructure
  elements
• Attack
• Systematic failure of nodes and/or infrastructure
  elements
• Scale-free networks are failure-tolerance
• Why?
• Most P2P systems give priority for
  failure-tolerance over attack-tolerance

31
Possible Targets

• Underlying protocol layers
• P2P routing mechanism
• Nodes themselves
• Trust system
• Homeostasis (of the system)
• Applications/Application Protocols
• Users

32

• Still not sure what to write here ?
• Anyways
• Questions?