Brief description on P2P networks and then some slides about PeerSim simulation on Max Min observer functions.

1. Web and Social Computing
Speakers: Chandan ,Robert and Kalthoom

2.  Objectives
 Peer to Peer Networks
 PeerSim
 Max and Min function Implementation and Results
 Conclusion
 Question & Answers

3. 1. Describe PeerSim functionality
2. Describe the different modes of PeerSim .
3. Design new methods ( Maximum and Minimum ).
4. Explain the out put results .

4.  A peer-to-peer network, P2P, is considered to be a
type of network architecture composed of
contributors that make their resources available to
other contributors on the same network without the
need for a server to oversee the transfer of
information.

5.  A peer-to-peer network’s architecture is generally formed dynamically by an ad-
hoc type of addition of nodes, meaning the loss of a certain node does not have a
significant impact on the rest of the network.
 This aspect of peer-to-peer networks gives the system the ability to easily scale to
any size that may be needed for a certain application.

6.  There are generally two major
types, Structured and
Unstructured. These two types of
networks also contain even more
specific underlying types of
networks.

7.  To be considered an unstructured
peer-to-peer network, the network
must not use any algorithm for
organization or optimization of
the network.
 For unstructured peer-to-peer
networks there are three different
models which include Pure peer-
to-peer networks, Hybrid peer-to-
peer networks, and Centralized
peer-to-peer networks.

8.  In a pure peer-to-peer network, or sometimes called a purely decentralized
network, all nodes are of equal ability meaning that there are no nodes that have
any special infrastructure function that could affect the network.
 A great example of a popular purely decentralized peer-to-peer network is the
Gnutella network. The main purpose of the Gnutella network is for that of a file
sharing system.

9.  Infrastructural nodes are
allowed to exist and are
often a type of central
directory server.
 All clients connected to the
network must connect to
one of these servers.
 Users connection
information and List of files
are saved in the central
server.

10.  Added Supernodes.
 These Supernodes are nodes
that are dynamically assigned
the task of servicing a small
subpart of the peer network by
indexing and caching files
contained therein.
 A popular example of this
centralized peer-to-peer
network is the file sharing
network Kazaa.

11.  Structured peer-to-peer networks generally hold the traits of using some type of
algorithm for organization or optimization of the network.
 Structured peer-to-peer networks use some type of global protocol to ensure that
any node in the network can proficiently route a search to some peer that has a
desired file.
 By far the most popular type of structured peer-to-peer network is the distributed
hash table or DHT.

12. • N nodes in the network
• Consistent hashing is used to assign keys and
values to nodes and resources
- SHA-1 hash of node’s IP address produces
160-bit ID
- SHA-1 hash of file produces 160-bit key, k
• Resource with key K is stored at node with ID=k
- If node with ID=k doesn’t exist then resource
is stored in the node with next highest available ID
• If node n joins the network
- Reassign keys from successor(n)
• If node n leaves the network
- Reassign keys to successor(n)

14.  Supplier vs Consumer Concept.
 Unlike Client/Server networks peers in P2P don’t rely one single source for
acquiring their necessities. Multiple Participants are involved.
 P2P is relatively cheaper to setup and simpler than client server networks.
 P2P can operate on a basic PC operating system whereas Client/ Server networks
require special OS.

15.  One of the major advantages to a peer-to-peer network is that each time a new
node is connected to the network the total capacity of the system increases.
 This point differs from the client-server model standpoint since the addition of
new nodes to a network that connects to a set of number of servers could possibly
slow data transfers for all of the connected users.
 Another advantage in regards to a pure peer-to-peer network is that there is no
single point of failure in the network.
 When compared to the client-server model on this point if the server fails the
network is brought down since the server is unable to relay information to the
connected nodes.

16.  One of the weaknesses of peer-to-peer networks is that of security. Generally
speaking peer-to-peer networks are more susceptible to security problems.
 Another disadvantage of peer-to-peer networks is the high bandwidth usage
required. This high bandwidth usage has lead to Internet Service Providers
starting to throttle, or limit, peer-to-peer traffic

17.  Simulating Peer-to-Peer (P2P) overlay networks is a common problem for
researchers and developers:
 need to be scalable .
 are highly dynamic .
 PeerSim :is one of the most known among researchers.

18.  A Peer-to-Peer simulator
 Is to use a modular approach, as the preferred way of coding with it is to re-use
existing modules:
 construct and initialize the underlying network modules.
 handle the different protocols modules.
 control and modify the network modules.
 The engines consist of components which may be 'plugged in' .

19.  Two different modes:
 Cycle-based :based on a very simple time scheduling algorithm and is very efficient and
scalable.
 Event-based :scheduled through events .

20.  protocols: used to define the behavior of the different peers.
 nodes: represent the peer themselves in the P2P network.
 controls: control the simulation by perform the global initialization and analysis:
 Initializer
 Dynamic
 Observers
 Linkable :interface used to access and manage node’s properties
 Add neighbor
 Get neighbor
 Node’s degree

21.  Every simulation is
configured with the help
of a configuration file :a
simple ASCII file

22. Simulator
• Main() Method
CDSimulator
nextExperment()
• loadInitializers()
• loadControls()
• execute()
Controls
Eg:
AverageObserver
s
Protocol n
IncrementalStats
Max, Min, Nmax,
Nmin, AVG etc
Protocol1 Protocol 2
Config File
• Set of Protocols and ordering
• Set of Controls and Ordering
• Set of Initializers + ordering
AverageFunction
• nextCycle()

23. Generated in Eclipse by ObjectAid UML Explorer

24.  How much free space does the network have to offer?
 What is the average lifetime of a node in the network?
 How many documents do nodes share on the average?
 Which node(s) performed the most searches?
 What is the maximum number of downloads that some node supported during a
session
 Etc
MAYANK BAWA, HECTOR GARCIA-MOLINA, ARISTIDES GIONIS, RAJEEV MOTWANI
“Estimating Aggregates on a Peer-to-Peer Network”
Stanford University, CA 94305 USA

26.  AverageObserver
 IncrementalStats

27.  Config File  Output: The Min Function Remains Constants and
the Max Values decreases towards the Min
value

28.  Config File  Results: Max Remains the same and Min increases
towards the Max Value
PeerSim Output

29.  Data taken in Excel  Graph generated

30.  Graph Generated Data taken in Excel

31.  Lessons Learnt (slide)
 Clear understanding of how Peer to Peer Networks work
 PeerSim provides an easy way to simulate P2P Protocols without wasting resources
 Java Programming Revised.
 Limitations
 The PeerSim Source codes are somehow complex to understand
 Some Protocols implemented in PeerSim need some time to understand

32. Q& A