|
Notes:
|
Peer-to-Peer (P2P) computing, the harnessing of idle CPU cycles through the Internet, offers
new research challenges in distributed computing. One of the most critical aspects in the
design of P2P systems is the development of incentive techniques to encourage cooperation
and resource sharing among participants. To tackle this problem, in this paper we propose
an incentive and scheduling mechanism based on credits with a two-level topology
which is designed to operate on different types of shared computing networks, such as
P2P, P2P Grid, Opportunistic Grid, Desktop Grid, volunteer computing platforms, among
others. The low level association of peers is made up of one super-peer and its constituent
peers. The main contribution at this level is a new reinvestment policy, called ‘‘Weighted’’,
that increases peer participation significantly. This mechanism reflects P2P user dynamics,
penalizes free-riders efficiently and encourages peer participation. Scalability limitations
on the first level are avoided by providing the mechanism with an upper level, made up
of super-peers interconnected through an upper level overlay. Our main contribution at
this level is the proposal of a task scheduling policy based on three criteria, computing
capacity with neighbors, distance and reputation that only use local information for mapping
tasks. Simulation results show our proposal performs well and is functional.
This work was supported by the MEyC-Spain under contract TIN2011-28689-C02-02 and CSD-2007-00050 and the European Social Fund. The authors are members of the research group 2009SGR145, funded by the Generalitat de Catalunya. |
