Two phase approach for grid enablement of legacy applications

Abstract

Grid enabling of existing applications has been a prime area of research in recent times and numerous solutions have been proposed in this regard. However, they are mostly tailor-made for specific application(s)/application scenarios and require significant domain knowledge and programming effort. In this paper, we propose a two phase approach for efficient grid enablement of legacy applications with minimal manual intervention. We present an analysis of generic application architecture, and the subsequent process of application profiling and mapping of programming components to this generic architecture for effective grid enablement. We discuss the stages of application partitioning, enablement and deployment. We illustrate the same in the case of a C application.Index Terms-grid computing, grid enabling of applications.

Grid Enablement of Applications Using Application Dependency Graph

“…In this section, we provide a brief overview of our proposed two phase approach to grid enablement [1]. Fig.…”

“…In our previous work [1] we have proposed a two phase approach for efficient grid enablement, wherein the application is extensively profiled, partitioned and grid enabled as a set of well coupled grid jobs and services. Determining the optimum partitioning of an application is the cornerstone for efficient grid enablement of the application.…”

“…Determining the optimum partitioning of an application is the cornerstone for efficient grid enablement of the application. In [1], we have detailed the mapping of different programming components to a generic application architecture [12]. A holistic view of the components of the application is obtained in the form of an Application Dependency Graph (ADG), which is also independent of the programming language.…”

“…In our previous work [1] we have proposed a two phase approach for efficient grid enablement of legacy applications with minimal manual intervention. The first phase in this approach results the generation of a graph that represents various components of the application as nodes and the dependencies between them as directed edges.…”

“…The ADG presents a low level of abstraction of the application, and will be considerably large for even simple applications. A holistic approach in addressing dependencies as well as scaling them is required for efficient grid enabling and deployment of the application, so that the nodes of the resulting ADG can be mapped to corresponding grid jobs/services, which are integrated through a client application [1] that represents the main control of the grid enabled application.…”