Data replication is a means of increasing the availability, scalability, and performance of data grids. It involves two main tasks: replica placement and replica selection. The goal of replica placement is to choose the optimal node for placing a copy of the data, whereas replica selection targets the choice of the optimal node for serving the user request.

The authors provide a thorough survey of existing replication approaches for data grids. They present a very helpful categorization of existing replica placement and selection techniques, together with the evaluation parameters and the performance results of the different techniques. This makes the paper a good basis for understanding the different replication approaches and the consequences of using one or the other.

According to the authors, there are many different conflicting optimization parameters, such as performance, cost, and so on, that influence both replica placement and selection. Existing techniques target different subsets of these parameters, making their comparison very difficult. This paper provides sufficient information for comparing existing techniques with each other and for mapping the development of new approaches to existing ones.

It is definitely worth reading, not only for those interested in getting a detailed overview of existing replication approaches, but also for those conducting research in the area of data replication.