EBC: Application-level migration on multi-site cloud

Cloud computing providers offer their services that help IT organizations to increase hardware utilization dramatically, and to scale up computational infrastructure instantly. The cloud resource management system could suspend/migrate some running jobs/virtual machines in order to reserve resources for other ones to improve resource utilization. In this case, the cost of moving the whole virtual machines is too high due to the big size of virtual machines. Moreover, supporting communication between the two migrated virtual machines is still a hard problem. We are interested in using checkpointing and recovery technique (C&R) to stop and restart running programs on virtual machines. The paper introduces the event-based checkpointing tool (EBC), in which users can easily checkpoint and restart running programs in cloud systems. EBC is based on event-driven architecture and so it is integrated into cloud infrastructures like OpenNebula easily. EBC supports not only sequential programs but also MPI programs. Moreover, any C&R included in MPI libraries is reused in EBC and so it is independent from MPI library versions. EBC is a useful tool supporting migration as well as scheduling in cloud systems.