Failure Analysis of Hadoop Schedulers using an Integration of Model Checking and Simulation

The Hadoop scheduler is a centerpiece of Hadoop, the leading processing framework for data-intensive applications in cloud. Given impact failures on performance running testing and verifying critical. Existing approaches such as simulation analytical modeling are inadequate because they not able to ascertain complete verification scheduler. This due wide range constraints aspects involved Hadoop. In this paper, we propose novel methodology that integrates combines model checking techniques perform formal schedulers, focusing following properties: schedulability, fairness resources-deadlock freeness. We use CSP language formally describe scheduler, PAT checker verify its properties. Next, proposed analyze OpenCloud, Hadoop-based cluster simulates load, order illustrate usability benefits our work. Results show can help identify several tasks (up 78%) early on, i.e., before executed cluster.