The Peer-to-Peer (P2P) architectures that are most prevalent in today's Internet are decentralized and unstructured. Search is blind in that it is independent of the query and is thus not more effective than probing randomly chosen peers. One technique to improve the effectiveness of blind search is to proactively replicate data. We evaluate and compare different replication strategies and reveal interesting structure: Two very common but very different replication strategies - uniform and proportional - yield the same average performance on successful queries, and are in fact worse than any replication strategy which lies between them. The optimal strategy lies between the two and can be achieved by simple distributed algorithms. These fundamental results o.er a new understanding of replication and show that currently deployed replication strategies are far from optimal and that optimal replication is attainable by protocols that resemble existing ones in simplicity and operation.
[1]
Ben Y. Zhao,et al.
An Infrastructure for Fault-tolerant Wide-area Location and Routing
,
2001
.
[2]
Antony I. T. Rowstron,et al.
Storage management and caching in PAST, a large-scale, persistent peer-to-peer storage utility
,
2001,
SOSP.
[3]
Wushow Chou,et al.
Queueing Systems, Volume II: Computer Applications - Leonard Kleinrock
,
1977,
IEEE Transactions on Communications.
[4]
Mark Handley,et al.
A scalable content-addressable network
,
2001,
SIGCOMM '01.
[5]
Edith Cohen,et al.
Search and replication in unstructured peer-to-peer networks
,
2002,
ICS '02.
[6]
Jussi Kangasharju,et al.
Optimal content replication in P2P communities
,
2002
.
[7]
Nitin H. Vaidya,et al.
Scheduling data broadcast in asymmetric communication environments
,
1999,
Wirel. Networks.