GridLab
Grid Application Toolkit

A simple API for Grid Applications
GAT

Menu



next up previous contents
Next: Cloning Jobs Up: Job Management Previous: Stopping Jobs   Contents

Checkpointing Jobs

Before telling you how to ``checkpoint'' I'll take a step back and tell you what ``checkpointing'' is. Checkpointing is the process of saving the state of a job to a long term storage medium. In the future, at least the future according to Ray Kruzwell, people will be able to save the contents of their brains to hard drives, or whatever fancy thing takes the place of hard drives then. If you were to do so, then all you know, all you are, would be etched on these little spinning platers in one's and zero's. It would be like some type of insurance policy that would allow you to walk out the door and get hit by a car, but then be reconstituted from the little one's and zero's on this spinning platter. The process of writing all that is you to these hard drives you be the process of checkpointing you. It would allow you to continue in the knowledge that if something were to go wrong, you could be restored to the you that existed just when you dumped your brain to disk.

Checkpointing a Job is easy, much easier then checkpointing you brain and much easier then stopping the crime-fighting onslaught that was ``The Brown Hornet.'' One simply calls the Checkpoint function on a Job. All of the instrumentation involved with determining how to checkpoint the job is handled by the magic that is GAT.


next up previous contents
Next: Cloning Jobs Up: Job Management Previous: Stopping Jobs   Contents
Andre Merzky 2004-05-13