Peer to Peer Computing

1
Peer to Peer Computing

• from
• http//www.cs.rpi.edu/courses/fall02/netprog/

2
What is Peer-to-Peer?

• A model of communication where every node in the
  network acts alike.
• As opposed to the Client-Server model, where one
  node provides services and other nodes use the
  services.

3
Advantages of P2P Computing

• No central point of failure
• E.g., the Internet and the Web do not have a
  central point of failure.
• Most internet and web services use the
  client-server model (e.g. HTTP), so a specific
  service does have a central point of failure.
• Scalability
• Since every peer is alike, it is possible to add
  more peers to the system and scale to larger
  networks.

4
Disadvantages of P2P Computing

• Decentralized coordination
• How to keep global state consistent?
• Need for distributed coherency protocols.
• All nodes are not created equal.
• Computing power, bandwidth have an impact on
  overall performance.
• Programmability
• As a corollary of decentralized coordination.

5
P2P Computing Applications

• File sharing
• Process sharing
• Collaborative environments

6
P2P File Sharing Applications

• Improves data availability
• Replication to compensate for failures.
• E.g., Napster, Gnutella (guh-noo-tell-ahh),
  Freenet, KaZaA (kuz-zaah).

7
P2P Process Sharing Applications

• For large-scale computations
• Data analysis, data mining, scientific computing
• E.g., SETI_at_Home(search for extraterrestrial
  intelligence), Folding_at_Home, distributed.net,
  World-Wide Computer

8
P2P Collaborative Applications

• For remote real-time human collaboration.
• Instant messaging, virtual meetings, shared
  whiteboards, teleconferencing, tele-presence.
• E.g., talk, IRC, ICQ, AOL Messenger, Yahoo!
  Messenger, Jabber, MS Netmeeting, NCSA Habanero,
  Games

9
P2P Technical Challenges

• Peer identification
• Routing protocols
• Network topologies
• Peer discovery
• Communication/coordination protocols
• Quality of service
• Security
• Fine-grained resource management

10
P2P Topologies

• Centralized
• Ring
• Hierarchical
• Decentralized
• Hybrid

11
Centralized Topology
Napster central indexing and searching service
12
Ring Topology
13
Hierarchical Topology
14
Decentralized Topology
Gnutella, Freenet, OceanStore
15
Hybrid TopologyCentralized Ring
16
Hybrid TopologyCentralized Decentralized
17
Evaluating topologies

• Manageability
• How hard is it to keep working?
• Information coherence
• How authoritative is info? (Auditing,
  non-repudiation)
• Extensibility
• How easy is it to grow?
• Fault tolerance
• How well can it handle failures?

18
Evaluating topologies

• Resistance to legal or political intervention
• How hard is it to shut down? (Can be good or bad)
• Security
• How hard is it to subvert?
• Scalability
• How big can it grow?

19
Decentralized

• Manageable
• Coherent
• Extensible
• Fault Tolerant
• Secure
• Lawsuit-proof
• Scalable

• Very difficult, many owners
• Difficult, unreliable peers
• Anyone can join in!
• Redundancy
• Difficult, open research
• No one to sue
• Theory yes Practice no

20
Centralized Decentralized

• Manageable
• Coherent
• Extensible
• Fault Tolerant
• Secure
• Lawsuit-proof
• Scalable

• Same as decentralized
• Better than decentralized
• Anyone can still join!
• Plenty of redundancy
• Same as decentralized
• Still no one to sue
• Looking very hopeful

Best architecture for P2P networks?
21
Napster

• The P2P revolution is started.
• Central indexing and searching service
• File downloading in a peer-to-peer point-to-point
  manner.

22
Gnutella

• Peer-to-peer indexing and searching service.
• Peer-to-peer point-to-point file downloading
  using HTTP.
• A gnutella node needs a server (or a set of
  servers) to start-up gnutellahosts.com
  provides a service with reliable initial
  connection points

But introduces a new single point of failure!
23
Freenet

• Peer-to-peer indexing and searching service.
• Peer-to-peer file downloading.
• Files served use the same route as searches (not
  point-to-point)
• Provides for anonymity (freedom of speech)
• Communications are encrypted and are
  "routed-through" other nodes to make it extremely
  difficult to determine who is requesting the
  information and what its content is.

24
KaZaA/Morpheus

• Hybrid indexing/searching model
• Not centralized like Napster, not decentralized
  like Gnutella.
• Peer-to-peer file downloading using HTTP.
• SmartStream automatic download resumption (for
  incomplete file)
• FastStream fast downloads (distribute the
  download task over a list of peers)
• SuperNodes elected dynamically if sufficient
  bandwidth and processing power hybrid topology
  model.
• A central server keeps user registrations, logs
  usage, and helps bootstrapping peer discovery.

25
References

• Nelson Minars articles at
• http//www.openp2p.com/pub/a/p2p/2001/12/14/topolo
  gies_one.html
• http//www.openp2p.com/pub/a/p2p/2002/01/08/p2p_to
  pologies_pt2.html

26
Grid v.s. P2P?

• Refer to
• I. Foster and A. Iamnitchi. On death, taxes, and
  the convergence of peer-to-peer and grid
  computing. In Proceedings of the 2nd
  International Workshop on Peer-to-Peer Systems
  (IPTPS '03), 2003.

27
Towards High Performance Peer-to-Peer Content and
Resource Sharing Systems
28
Question No. 1

• Which type of P2P applications the paper
  addresses?
• File sharing
• Process sharing
• Collaborative environments

29
Question No. 2

• What is the goal of the paper?
• Ensure the users to access all available content
  efficiently
• Fair load distribution to ensure load balancing,
  considering the heterogeneity of the contributed
  resources by the different peers
• Low user-request response times

30
Question No. 3

• What are the challenges addressed by the paper in
  achieving the goal?
• Autonomous nodes ? complex distributed
  coordination algorithms
• Scalability in number of nodes, documents
• Heterogeneity in content contributions,
  processing and storage capacities
• Dynamism in content popularity, nodes, content

31
Question No. 4

• What are the assumptions made by the paper?
• Content has an initially static and known
  popularity, with document popularities following
  the Zipf distribution the popularity of any
  document is roughly inversely proportional to its
  rank in the popularity table,
• fi K/ia
• Who has the control on the node content?
• By the node user? (free-rider)
• By the Max-Fair algorithm?

32
Question No. 5

• How does the query works?
• Load-balancing
• intra-cluster
• inter-cluster
• associating the document categories with clusters
  of nodes, in a manner that ensures a fair
  distribution of the document-category
  popularities to the clusters of nodes
• fairness index

33
Max-Fair Algorithm

• How are clusters initially decided?
• What leads to a nodes belonging to more than one
  cluster?

34
Discussions

• Does the paper address the challenges well and
  achieve its goal?
• Fair load distribution to ensure load balancing,
  considering the heterogeneity of the contributed
  resources by the different peers
• Low user-request response times
• Deficiencies of the proposed strategy?

35
A Course Project

• Research on the state-of-art of load-balancing in
  Peer-to-Peer computing
• Propose an improved strategy by
• removing one deficiency or
• dropping one simplifying assumption
• in existing mechanisms
• Evaluate the proposed strategy by simulation