Dynamic QoS-aware data replication in grid environments based on data “importance”

Data replication comprises a standard fault tolerance approach for systems–especially large-scale ones–that store and provide data over wide geographical and administrative areas. The major topics that the task of data replication covers include the replica creation, placement, relocation and retirement, replica consistency and replica access. In a business context a number of constraints exists which are set by the infrastructure, network and application capabilities in combination with the Quality of Service (QoS) requirements that hinder the effectiveness of data replication schemes. In this paper, we examine how this combination affects the replication lifecycle in Data Grids and we introduce a set of interoperable novel file replication algorithms that take into account the infrastructural constraints as well as the ‘importance’ of the data. The latter is approximated through a multi-parametric factor that encapsulates a set of data-specific parameters, such as popularity and content significance.

Research highlights► Introduction of the data “importance” concept in the data replication context. ► Development of a set of mechanisms for QoS-aware replica management, encapsulating a set of data-specific parameters, such as popularity and content significance. ► Reduction of the replica management problem into a search problem allowing for efficient approaches for its solving.