170 likes | 452 Views
SELF-CHORD. a Bio-Inspired P2P Framework for Self-Organizing Distributed Systems Agostino Forestiero Emilio Leonardi Carlo Mastroianni Michela Meo Sonal Srivastava. P2P Models. Unstructured. Structured. Better network management L ow efficiency
E N D
SELF-CHORD a Bio-Inspired P2P Framework for Self-Organizing Distributed Systems AgostinoForestiero Emilio Leonardi Carlo Mastroianni MichelaMeo SonalSrivastava
P2PModels Unstructured Structured • Better network management • Low efficiency • of discovery procedures • More efficient in • terms of search time and network load • Can limit the expressiveness of discovery requests— users are only allowed to search for specific resources, • but cannot issue complex or “range” queries.
Objective • “To show how the most popular structured system, Chord, can be enriched with self-organization and adaptive properties.”
Structured topology of CHORD SELF-CHORD Self-organizing mechanisms for key ordering and discovery
CHORD • Each peer is assigned a binary code, or “key,” by a hash function • Peers are organized in a ring and ordered following the values of their keys. • Resources are also indexed by keys • Each resource is consigned to the peer that has the same key as the resource or, if such a peer is not present in the ring, on the first following peer, also called “successor.”
How are self-organizing properties obtained in Self-Chord? • By the use of mobile agents, which travel the Chord ring and order resource keys • Mobile agents operations are inspired by the behaviour of ants, in accordance with the swarm intelligence paradigm
Ant Algorithms • one of the most popular examples of “swarm intelligence” systems. • a number of agents follow very simple rules with no centralized control • complex global behavior emerges from their local interactions.
Basic features of Self-Chord • peers are organized and ordered in a ring according to their indexes, like in Chord. • Differently from Chord, however, resource keys are not strictly correlated with peers. • A key is not consigned to the peer whose index is the same as the key, or to its successor (the first peer on the network with a higher index).
Reorganization of keys in Self-Chord • For each peer, its centroid is defined. • The agents try to deposit each key in the peer whose centroid in as close as possible to the key • Probabilistic procedures are defined, so that the mobile agents tend to pick a key from a peer if the key is an outlier there, and try to move the key to a peer whose centroid is similar to the key value
Benefits • possibility to serve range queries • load balancing • dynamic behavior
Scalability Analysis The value of the large majority of the keys is very close to the peer centroid
Performance of Discovery Requests the discovery procedure successfully discovers nearly all the keys that have the desired value.
Related Work • Anthill • So-Grid • Antares
Conclusion • This paper aims to open a new research avenue for P2P frameworks • it presents a P2P system that inherits the beneficial characteristics of structured systems, but offers further profitable characteristics inherited by biological systems
Future Work • Ant inspired approach can be applied to other structured P2P systems in which peers are not organized in a ring, but in other structures such as multidimensional grids or trees.