GREEDY SINGLE USER AND FAIR MULTIPLE USERS' REPLICA SELECTION DECISION IN DATA GRID

Replication in data grids increases data availability, accessibility and reliability.
Replicas of datasets are usually distributed to different sites, and the choice of any
replica locations has a significant impact. Replica selection algorithms decide the best
replica places based on some criteria. To this end, a family of efficient replica
selection systems has been proposed (RsDGrid). The problem presented in this thesis
is how to select the best replica location that achieve less time, higher QoS,
consistency with users' preferences and almost equal users' satisfactions. RsDGrid
consists of three systems: A-system, D-system, and M-system. Each of them has its
own scope and specifications. RsDGrid switches among these systems according to
the decision maker. The first two systems improve the existing single user replica
selection systems by incorporating new features, while M-system presents a new
direction focusing on simultaneous multiple users' selection.