Designing a Dependable Job Submission System in gLite
Filippo Cuttone, Antonella Di Stefano, Giovanni Morana · 2010
The high level of abstraction and the increasing ease of use achieved by middleware have certainly contributed to the diffusion of the grid computing paradigm. However, in order to obtain transparency and simplicity towards users it was necessary to complicate both the internal layout and the organization of the middlewares. Today, a grid middleware is composed of numerous software services that, interacting among them, manage many tasks, which are not always independent of each other: if it is not properly monitored and managed, this type of organization can lead to errors in middleware functioning that directly affect on overall system performance. The solution taken by current implementations of the main grid middlewares to reduce the impact of such errors are inadequate: they are limited to a small number of very common errors and usually consist of the repetition of step failed. In this paper, the authors take into account the issues related to the dependability in grid middleware and give an example of a dependable job submission process using the gLite middleware.