bio inspired peer to peer systems

download bio inspired peer to peer systems

of 32

Transcript of bio inspired peer to peer systems

  • 7/30/2019 bio inspired peer to peer systems

    1/32

    P2P Systems - 1

    Bio Inspired Peer-to-Peer Systems(Inspired by Ant Colonies)

    Research Paper - RAFFAELE GIORDANELLI andCARLO MASTROIANNI, ICAR-CNR

    and eco4cloud srl, Italy

    MICHELA MEO, Politecnico di Torino, Italy

    Navneet Gupta109403CO - 3

  • 7/30/2019 bio inspired peer to peer systems

    2/32

    P2P Systems - 2

    Content

    Centralized Information Systems

    (Semi-)Decentralized Information Systems Lessons Learned from Napster

    Fully Decentralized Information Systems

    Self-Organization

    Resource Location in P2P Systems

    Unstructured P2P Overlay Networks Protocol Message Types

    Hierarchical P2P Overlay Networks

    Structured P2P Overlay Networks A GENERALIZED ANT-INSPIRED STRATEGY FOR STRUCTURED P2P

    SYSTEMS Algorithm

    Description

    Multi Dimensional Overlay

    Conclusion

  • 7/30/2019 bio inspired peer to peer systems

    3/32

    P2P Systems - 3

    Web search engine Global scale application

    Example: Google

    Centralized Information Systems

    Google: 15000 servers

    1

    Find"aberer"

    2

    Resulthome page of Karl Aberer

    GoogleServer

    Client

    Client

    Client

    ClientClient

    Client

    Client

    Client

    Client

    Client

    Strengths Fast response time

    Weaknesses Infrastructure, administration, cost

    A new company for every global application ?

  • 7/30/2019 bio inspired peer to peer systems

    4/32

    P2P Systems - 4

    (Semi-)Decentralized Information Systems

    P2P Music file sharing Global scale application

    Example: Napster 1.57 Mio. Users

    10 TeraByte of data(2 Mio songs, 220 songs per user)(February 2001)

    1

    Find

    "brick in the wall" "pink floyd" "1 MB" "rock" schema

    2

    Resultyou find f.mp3 at peer x

    3

    Request and transfer filef.mp3from peer X directly

    NapsterServer

    Peer

    Peer

    Peer

    PeerPeer

    Peer

    Peer

    Peer

    Peer

    PeerX

    Napster: 100 servers

  • 7/30/2019 bio inspired peer to peer systems

    5/32

    P2P Systems - 5

    Lessons Learned from Napster

    Strengths: Resource Sharing Every node pays its participation by providing access to its resources

    physical resources (disk, network), knowledge (annotations), ownership (files)

    Every participating node acts as both a client and a server : P2P

    global information system without huge investment

    decentralization of cost and administration = avoiding resource bottlenecks

    Weaknesses: Centralization server is single point of failure

    copying copyrighted material made Napster target of legal attack

    Centralized

    System

    Decentralized

    System

    increasing degree of resource sharing and decentralization

  • 7/30/2019 bio inspired peer to peer systems

    6/32

    P2P Systems - 6

    Fully Decentralized Information Systems

    P2P file sharing Global scale application

    Example: Gnutella 40.000 nodes, 3 Mio files

    (August 2000)

    Gnutella: no servers

    Strengths No infrastructure, no administration

    No single point of failure

    Weaknesses High network traffic

    No structured search

    Free-riding

    Find"brick in the wall"

    I have"brick_in_the_wall.mp3".

    Self-organizing System

  • 7/30/2019 bio inspired peer to peer systems

    7/32

    P2P Systems - 7

    Self-Organization

    Self-organized systems well known from physics, biology, cybernetics distribution of control ( = decentralization = symmetry in roles = P2P) local interactions, information and decisions

    emergence of global structures

    failure resilience

    Each component acts locally and autonomously based on local information on

    the state of the system. Self-organizing systems exhibit structures andstructured behavior, that are not enforced by some outside entity, but thatemerge from the aggregation of the local interactions of the components ofthe systems

  • 7/30/2019 bio inspired peer to peer systems

    8/32

    P2P Systems - 8

    Resource Location in P2P Systems

    Problem: Peers need to locate distributed information Peers with address p store data items d that are identified by a key kd Given a key kd (or a predicate on kd) locate a peer that stores d,

    i.e. locate the index information(kd, p)

    Thus, the data we have to manage consists of the key-value pairs (kd, p)

    Can such a distributed database be maintained and accessed by a set ofpeers without central control ?

    P1

    P2 P3

    P4

    P5

    P6P7

    P8kd="xyz"p="P8"

    d=xyz.docx"

    kd ="xyz" ?

    (xyz",P8)

  • 7/30/2019 bio inspired peer to peer systems

    9/32

    P2P Systems - 9

    1. Unstructured P2P Overlay Networks

    No index information is used i.e. the information (k, p) is only available directly from p

    Simplest approach: Message Flooding (Gossiping) send query message to C neighbors

    messages have limited time-to-live TTL

    messages have IDs to eliminate cycles

    k=xyz"

    Example: C=3, TTL=2

  • 7/30/2019 bio inspired peer to peer systems

    10/32

    P2P Systems - 10

    Protocol Message Types

    Type Description Contained Information

    Ping Announce availability and probe forother servents

    None

    Pong Response to a ping IP address and port# of responding servent;number and total kb of files shared

    Query Search request Minimum network bandwidth of respondingservent; search criteria

    QueryHit Returned by servents that havethe requested file

    IP address, port# and network bandwidth of

    responding servent; number of results and

    result set

    Push File download requests for

    servents behind a firewall

    Servent identifier; index of requested file; IP

    address and port to send file to

  • 7/30/2019 bio inspired peer to peer systems

    11/32

    P2P Systems - 11

    CA

    B D

    EAs pingBs pongCs pong

    Ds pong

    Es pong

  • 7/30/2019 bio inspired peer to peer systems

    12/32

    P2P Systems - 12

    Searching (Query/QueryHit/GET)

    CA

    B D

    EAs query (e.g., X.mp3)Cs query hit

    Es query hit

    X.mp3

    X.mp3

    GET X.mp3 X.mp3

  • 7/30/2019 bio inspired peer to peer systems

    13/32

    P2P Systems - 13

    2. Hierarchical P2P Overlay Networks

    Servers provide index information, i.e. the information (k, p) is availablefrom dedicated servers

    Simplest Approach one central server

    user register files

    service (file exchange) is organized

    as P2P architecture

    k="jingle-bells"

    index server

  • 7/30/2019 bio inspired peer to peer systems

    14/32

    P2P Systems - 14

    Napster

    Central (virtual) database which holds an index of offered MP3/WMAfiles

    Clients connect to this server, identify themselves (account) and send alist of MP3/WMA files they are sharing (C/S)

    Other clients can search the index and learn from which clients they canretrieve the file (P2P)

    Additional services at server (chat etc.)

    A B

    Napster Serverregister

    (user, files)

    Ahas X.wmv

    Download X.wmv

  • 7/30/2019 bio inspired peer to peer systems

    15/32

    P2P Systems - 15

    3. Structured P2P Overlay Networks

    Peers are organized in accordance with a predefined logical structure ring, multi-dimensional grid, tree and resources are assigned to peerswith a precise strategy.

    These are generally based on DHT, whose purpose is to assign keys toresources, and indexes to peers, by means of hash functions.

    Each peer is responsible for a portion of the key space, and is requiredto manage the keys belonging to this portion.

    However, the use of hash functions has an important drawback: since

    similar resources are mapped to completely different keys, they aregenerally assigned to peers that are located far from each other in theoverlay. This hinders the efficient execution of queries for resourceshaving similar characteristics, or range queries.

  • 7/30/2019 bio inspired peer to peer systems

    16/32

    P2P Systems - 16

    Distribution of Index Information

    Goal: provide efficient search using few messages without usingdesignated servers

    Easy: distribution of index information over all peers, i.e. every peermaintains and provides part of the index information (k, p)

    Difficult: distributing the data access structure to support efficientsearch

    index information I

    server

    data access

    structure

    peers (storing data)

    peers (storing data and index information)

    I1 I2 I3 I4

    ?

    Search starts here

    Where to start the search?

  • 7/30/2019 bio inspired peer to peer systems

    17/32

    P2P Systems - 17

    Topological Routing (CAN)

    Based on hashing of keys into a d-dimensional space (a torus) Each peer is responsible for keys of a subvolume of the space (a zone) Each peer stores the adresses of peers responsible for the neighboring

    zones for routing

    Search requests are greedily forwarded to the peers in the closest zones

    Assignment of peers to zones depends on a random selection made bythe peer

  • 7/30/2019 bio inspired peer to peer systems

    18/32

    P2P Systems - 18

    A GENERALIZED ANT-INSPIRED STRATEGY FORSTRUCTURED P2P SYSTEMS

    INSPIRATION

    In any kind of structured P2P systems, the main objective is to provideclients a way to rapidly discover the desired resources along thedistributed overlay.

    This is done exploiting the Distributed Hash Table paradigm: everyresource is given a key using a hash function applied to the resourcename, and the key is assigned to a node of the structure whose code isequal or very close to the resource key.

    The hash function is needed to fairly distribute the resources over thestructure, but this is possible only if the popularity of the resources isuniform. If this is not the case, some peers may be overloaded, sinceresources with the same name are mapped to the same peer.

  • 7/30/2019 bio inspired peer to peer systems

    19/32

    P2P Systems - 19

    Moreover, the keys of resources having similar features are inevitablydispersed. This prevents the possibility of executing a range queryefficiently, since the target resources are most likely to be located indifferent and remote regions of the network.

  • 7/30/2019 bio inspired peer to peer systems

    20/32

    P2P Systems - 20

    Use Of Swarm Intelligence

    The basic idea is to use resource attribute(s) directly as the key,without using any hash function, and adopt an ant-inspired algorithm tosort the keys over the structure and, at the same time, distribute themto the peers in a fair fashion.

    This implies that resource keys are decoupled and independent from

    peer codes, while the two entities are strictly correlated in classical P2Psystems.

  • 7/30/2019 bio inspired peer to peer systems

    21/32

    P2P Systems - 21

    PREREQUISITES

    First, it is necessary to choose the overlay that best fits the type ofattributes used for indexing and searching the resources.

    The second step is to define, for each peer, its centroid. The peercentroid must represent, with a single quantity, the set of keys storedin a local region of the structure that comprises the peer itself and a

    few close-by peers.

    Finally, a set of mobile agents must be generated to perform thereorganization. The agents use the underlying structure to travel thesystem, but do not alter the structure and the way it is managed.

  • 7/30/2019 bio inspired peer to peer systems

    22/32

    P2P Systems - 22

    ALGORITHM

    The agent can either be loaded, when it is moving a key between peers,or unloaded.

    When unloaded, the agent travels the P2P overlay structure performinga random walk through neighbouring peers. Whenever the agent visits anew peer, it attempts to pick a key from the peer.

    If a pick operation succeeds, the agent becomes loaded and starts aprocess aiming at dropping the key in a more appropriate peer.

    Thus, a loaded agent carrying a key k moves towards a region of the

    overlay structure that islikely populated with keys that are similar (orclose) to k.

    The agent moves makingjumps towards the desired region. Finally,whenever a loaded agent visits a new peer, it attempts to drop the key.

  • 7/30/2019 bio inspired peer to peer systems

    23/32

    P2P Systems - 23

    Ants Swarming in a Ring: the Self-Chord System

    Self-Chord is a structured P2P systems, uses the ring-shaped overlay toestablish P2P interconnections, but distributes resource keys amongpeers using the operations of ant-inspired mobile agents.

    A mobile agent arrives at a peer with centroid C=30, picks key 55, andthen hops to a region of the ring where peers are supposed to havecentroids close to 55.

  • 7/30/2019 bio inspired peer to peer systems

    24/32

    P2P Systems - 24

    Four macro operations are identified:Random Walk, Pick, Jump and Drop.

    Random Walk - While the agent is unloaded (does not carry any key), ittravels the network randomly, exploiting the links towards other peersthat are provided by the specific overlay.

    Pick -The goal of this operation is to pick a key that, being far from thecentroid (i.e., different from the keys typically stored in the peer).

    Jump- Once the agent has picked a key, it jumps to a region of theoverlay wherethe key is supposed to better respect the currentsorting.

    Drop - The objective is to drop the carried key in a peer that holdssimilar keys.

  • 7/30/2019 bio inspired peer to peer systems

    25/32

    P2P Systems - 25

    DESCRIPTION

    The collective operations of agents sort both the centroids and the keysover the corresponding overlay.

    A notable improvement of the new strategy is that resource keys can begiven a semantic meaning, which implies that similar resources arestored in neighbour peers and range queries can be executed efficiently.

    Indeed, once a search message has found a key included in the targetrange specified in the query, all the other keys can be discovered veryrapidly, exploring the neighbour peers.

    This efficiency is very hard to obtain in classical P2P systems, becausethe keys of similar resources are spread by the hash function.

  • 7/30/2019 bio inspired peer to peer systems

    26/32

    P2P Systems - 26

    Each peer, at the time that it connects to the network, generates anagent with a probability Pgen.

    The average number of agents Na that circulatein the network at agiven instant of time is associated with the average number of peersconnected in the network at the same time, Np,

    Na= Np Pgen

    2006, Karl Aberer, Manfred Hauswirth - EPFL-IC, Laboratoire de systmes d'informations rpartis

  • 7/30/2019 bio inspired peer to peer systems

    27/32

    P2P Systems - 27

    The ant-inspired strategy has furtherinteresting properties

    it is self-organizing, as the assignment of keys is not predetermined butemerges from the combined behaviour of very simple agents;

    it is completely decentralized, since agent operations are exclusivelydriven by local information;

    it guarantees stability: once a correct sorting has been achieved, it isquickly recovered

    it ensures a fair load balance, even in the presence of non uniform keypopularity.

  • 7/30/2019 bio inspired peer to peer systems

    28/32

    P2P Systems - 28

    Ant Based Strategy Applied to Diff. Overlays

    RING OVERLAY If a resource can be characterized by the value of a single attribute.

    The local region centered at a peer includes the peer itself, number ofpeers on the left, and the same number of peers on the right.

    For this scenario, agent operations can be instantiated as follows.

    Random Walk. The unloaded agent travels from adjacent to adjacentpeer.

    Jump. The agent examines the peers and jumps to the one that is theclosest to the target peer.

    Pick and Drop. Pick and drop probability functions are, respectively,directly and inversely proportional to the distance between theconsidered key and the local centroid.

  • 7/30/2019 bio inspired peer to peer systems

    29/32

    P2P Systems - 29

    MultiDimensional Overlay

    A multidimensional overlay is preferable when a resource is convenientlycharacterized by several independent attributes.

    Example - a Grid/Cloud infrastructure in which the hosts arecharacterized by CPU,memory, and bandwidth.

    Agent operations are specialized as follows -

    Random Walk. The agent travels from adjacent to adjacent peer,randomly choosing both the dimension and the direction.

    Pick. As usual, a key should be picked when its distance from the localcentroid is high.

    Jump. The agent moves along the dimension that allows to minimize the

    distancebetween the carried key and the centroid of the target peer.Drop. An agent that arrives at a new peer should drop the carried keyif its valueis similar to the peer centroid.

  • 7/30/2019 bio inspired peer to peer systems

    30/32

    P2P Systems - 30

    CONCLUSION

    This article has presented a generalized ant-inspired approach that canbe used to sort the resource keys over any kind of P2P overlay.

    The process sorting, performed through the pick and drop operations of

    mobile agents, is statistically driven, self-organising, and decentralized.

    This enables the possibility of preserving the values of significantresource attributes into the resource keys, instead of generating thelatter with hash functions.

    The advantages of this approach are numerous, ranging from a more

    efficient execution of complex queries to improved behaviour in termsof adaptivity and load balancing.

  • 7/30/2019 bio inspired peer to peer systems

    31/32

    P2P Systems - 31

    Refrences

    Bio-Inspired P2P Systems: The Case of Multidimensional Overlay byRAFFAELE GIORDANELLI and CARLO MASTROIANNI, ICAR-CNR andeco4cloud srl, Italy MICHELA MEO, Politecnico di Torino, Italy.

    Peer To Peer Networking :Configuring Issues and DistributedProgramming by ODD ARILDSKAFLESTAD and NINA KAUREL

  • 7/30/2019 bio inspired peer to peer systems

    32/32

    P2P Systems - 32

    Thank you!!

    Questions ?

    32