Distributed processing environments, such as grids, are one of the most important platforms for meeting the user's processing needs. These environments have the potential to meet the needs of users, but they also have their own problems, including the failure of the job
More
Distributed processing environments, such as grids, are one of the most important platforms for meeting the user's processing needs. These environments have the potential to meet the needs of users, but they also have their own problems, including the failure of the jobs. Several attempts have been made to overcome this problem, which in general can be divided into two categories of resource side methods and job side methods. All these methods need some kind of prediction of the resources or jobs status in order to pursue a proactive approach to failures. However, due to the dynamics of these environments, the developed models quickly lose their quality and thus can not effectively help with the methods mentioned. In this paper, first, by identifying the reasons for reducing the quality of predictors in the grid environment, a solution has been proposed to deal with it, and then the proposed solution has been applied in the context of job failures. The results of experiments on the two experimental environments of AuverGrid and Grid5000 showed that the proposed method would increase the quality by 0.02 and 0.06 respectively in these two environments.
Manuscript profile