130 likes | 226 Views
Search and Replication in Unstructured P2P Networks. Qin Lv, Pei Cao, Edith Cohen, Kai Li, Scott Shenker. Architectures for P2p Networks. Centralized : Systems have a constantly-updated directory hosted at central locations. E.g. Napster
E N D
Search and Replication in Unstructured P2P Networks Qin Lv, Pei Cao, Edith Cohen, Kai Li, Scott Shenker
Architectures for P2p Networks • Centralized: Systems have a constantly-updated directory hosted at central locations. E.g. Napster • Decentralized but structured: These systems have no central directory server, but they have a significant amount of structure. E.g. Freenet, etc. • Decentralized and Unstructured: E.g. Gnutella.
Problems • Unstructured P2P networks, like Gnutella, use the query method of flooding. • Such search mechanism is extremely unscalable, generating large loads on the network participants. • It is necessary to improve the search method.
Simulations • Four topologies are used for simulations: • Power-Law Random Graph (PLRG) • Normal Random Graph (Random) • Gnutella Graph (Gnutella) • Two-Dimensional Grid (Grid)
Methodology • Statistics are collected from the numPlace x numQuery queries. • For each object i, numPlace = 10; for each placement, numQuery = 100 • So there’re 1000 queries for each object.
Limitations of Flooding • TTL is used to control the number of hops that a query can be propagated.
Expanding Ring • It uses successive floods with increasing TTLs. • A node starts a flood with a small TTL. • If the search is not successful, the node increases the TTL and starts another flood. • This method is called expanding ring.
Perf. of Expanding Ring • Though expanding ring solves the TTL selection problem, it does not address the message duplication issue inherent in flooding.
Random Walks • It forwards a query message to a randomly chosen neighbor at each step until the object is found. • To reduce the delay, a requesting node sends k queries, and each query messages takes its own random walk. • Usually, 16 to 64 walkers give good results.
Replication Strategies • Owner replication: when a search is successful, the object is stored at the requester node only. • Path replication: the object is stored at all nodes along the path. • Random replication: randomly pick p of the nodes to store.
Summary • Several search algorithms and replication strategies for unstructured P2P networks are simulated and evaluated.