The unprecedented growth and increased importance of geographically
distributed data coupled with recent trends of globalization has
created a strong need for efficient sharing of data globally. In
this regard, P2P systems and GRIDs have revolutionized the world.
Incidentally, the performance of such systems is critical to user
satisfaction. Hence, we investigate efficient search and effective
load-balancing mechanisms (via data migration/replication) for P2P
systems and GRIDs. In particular, we are currently exploring issues
concerning the feasibility of sharing spatial data in P2P systems.
While much work has been done towards expediting search in file-sharing
P2P systems, issues concerning spatial indexing in P2P systems are
significantly more complicated due to overlaps between spatial objects
and the complexity of spatial queries. Incidentally, existing R-tree-based
structures for distributed environments (e.g., the MC-Rtree) are
not adequate for addressing the sheer scale, dynamism and heterogeneity
of P2P environments. Hence, we have proposed the P2PR-tree (Peer-to-Peer
R-tree), which is a new spatial index specifically designed for
P2P systems. The main features of P2PR-tree are two-fold. First,
it is hierarchical and performs efficient pruning of the search
space. Second, it is completely decentralized, scalable and robust
to peers joining/leaving the system. The main contributions of our
load-balancing technique for spatial P2P systems are as follows.
First, we propose a dynamic data placement strategy involving data
migration and replication, the objective being to provide load-balancing
and reduced query response times. Second, we propose a novel technique
for load-balancing. Our proposed techniques for P2P systems can
also be cost-effectively integrated into GRIDs. Notably, our performance
evaluations concerning search and load-balancing in spatial P2P
systems and in GRIDs have demonstrated the effectiveness of our
methods in reducing user response times significantly.
An overview of our proposal on the P2PR-tree can be found here(MS-PowerPoint),
while a summary of our contribution to load-balancing in GRIDs is
located here(MS-PowerPoint). If you are interested
in further details, please feel free to contact me at anirban@tkl.iis.u-tokyo.ac.jp.
My research website is http://www.tkl.iis.u-tokyo.ac.jp/~anirban
and my personal website is located at http://www.angelfire.com/dbz/anirban.
(Anirban Mondal)
|
|