50 likes | 66 Views
Explore efficient search algorithms like Controlled Flooding and Gossip in Peer-to-Peer networks to locate items without centralized servers. Learn about advantages, challenges, and enhancements like Random Walker for optimized searches.
E N D
Peer-to-Peer Search Algorithms Fundamental problem:search for a particular item in a network of peers without centralized servers Question:Where do you send your queries?
Controlled Flooding • Send the query for a particular item to known peer nodes (i.e., neighbours) • The query will be forwarded by these neighbours to their own neighbours • Until the query finds itself to the peer node hosting the item, or until a particular TTL (Time-to-Live)has been reached • Those who have seen the original query will not forward it again • The query will not be sent back to where it came from • controlled flooding
Controlled Flooding II. • Advantage: the search is fast • Since it is an expanded ring search • For a network of size n, it takes approximatelylog n hops to reach the item • Problem: too much message overhead for one search • Question: How do we save overhead?
Gossip • Modify the forwarding mechanism in controlled flooding • so that each peer node will only forward thequery with a certain probability • Advantage: reduced message overhead • Problem: may not guarantee that the item can be found, if it exists
Random Walker • Modify the query dissemination mechanism, so that the query is only sent to one of the neighbours, whois randomly chosen • Advantage: the overhead is further reduced • Problem: number of hops to reach the item is high, takes longer time to reach the item • The problem can be mitigated by using k parallel random walkers • rather than just one random walker • The problem can also be mitigated by replicating the item to more peer nodes • increasing its exposure torandom walkers