No abstract

No abstract

Desktop Clouds rely on volatile computing resources. For instance, platforms such as cuCloud and UnaCloud run scientific applications in virtual machines exploiting idle resources harvested in computer labs. Regretfully, these resources can be claimed by users, turned off and faulted at any time. The application running on these platforms suffer interference and interruptions that do not occur in dedicated platforms. We have been researching how to deal with these interruptions to increase the platform reliability and support applications running for large periods of time. This paper describes an application of our Global Snapshot Protocol, which can be employed for executing and pausing distributed applications running on desktop clouds. We found that, in these environments, the number of failures caused by desktop users is greater than the caused by hardware and communications. There, when a distributed system running in the virtual machines of a desktop cloud is paused, it can be restored in the same desktops, and successfully finish the application execution.