Virtualization aware job schedulers for checkpoint-restart

Application checkpoint and restart has been a widely studied problem over the last several decades. Despite immense volume of theory and several research project level implementations, there is very little by way of working solutions for the case of parallel distributed applications (such as MPI programs on a cluster). We describe our experiences in enhancing a job scheduler to leverage mechanisms of a virtual machine environment to support checkpoint-restart. We also describe the basic coordinated checkpoint-restart framework that we implemented on which this solution is based.