Analyzing the performance of volunteer computing for data intensive applications
S. Alonso Monsalve, Félix Garcı́a-Carballeira, Alejandro Calderón · 2016
Data intensive applications involve the processing of large datasets obtained from simulations or from large-scale experiments that are able to generate terabytes or petabytes of data. There are different computational solutions that use large-scale distributed systems for addressing the execution of data intensive applications: cluster-based and grid computing, cloud computing, and volunteer computing. Volunteer computing is a paradigm in which large numbers of computers, volunteered by members of the general public, provide computing and storage resources. In this paper, we analyze the usage and performance of BOINC, the main middleware system for volunteer computing, for data intensive applications. With this aim, we have developed a Complete simulator of BOINC infrastructures that takes into account all aspects present in the system: server, data servers, clients, scheduling, disks, and networks. The paper describes the simulator and presents the main simulation results obtained when a BOINC infrastructure is used for data intensive applications.