Modelling and Analysis of Various Grid Protocols in a Simple and General Manner

In the recent years, data replication schemes have been developed which use logical grid structures for managing the replicated data 4, 12, 19]. Grid structure-based protocols have been proven to be extremely eecient due to the high availability of read and write operations they ooer and due to the very small number of copies which have to be accessed during an operation execution. Although the diierent grid protocols use logical grids, their semantics and interpreting algorithms vary from scheme to scheme. This has the disadvantage that switching from one kind of grid protocol to another one, e.g. because of changing network or application characteristics , results in the re-implementation of the new protocol. But a well-known fact is that implementing and re-implementing replication schemes is extremely time-and money-consuming. Furthermore, these are highly error-prone tasks which must be consequently avoided. By designing a general framework which separates the policy from the mechanism these drawbacks can be overcome: a framework based on a canonical grid structure and a general algorithm for processing the grid conveniently allow the switch between diierent schemes simply by replacing the scheme-speciic grid structures. There is no need of modifying a single line of code. The purpose of this paper is to present such a general framework for grid protocols. The framework allows their exploration in a homogeneous and simple way. For this purpose, we show how to model popular grid protocols as instances of the framework, thereby demonstrating that grid protocols which look quite diierent at rst glance share a common concept of construction. Additionally, we explain how new grid protocols with various properties can be designed by using the framework in a graphical way.