The authors have managed to reduce the size and time of virtual machine (VM) checkpoints by over 60 percent. Their basic concept is to avoid writing memory pages to disk that are already there. While that’s easy to say, and may even seem obvious, it takes the authors six pages to present the details of their solution, and another three pages to explain their measurement methods and results.
At first glance, this would seem to be important only to those few virtualization software vendors who have to write backup tasks for their products. However, checkpoints are also used in live migrations--the moving of virtual machines from one physical host to another without shutting down. This research makes that process faster and more reliable, meaning it can be employed in more environments. How would you like to move a running application from your desktop to your phone to your automobile navigation computer?
Faster checkpoints also enable better security monitoring, such as taking continuous snapshots of attacks on virtual honeypots. Continuous snapshots would also provide a reliable recovery scheme for restoring crashed computers, or, on a larger scale, implementing a “time machine” (apologies to Apple) capability for applications and services in private and public clouds. Such dial-a-date selectivity could turn out to be crucial for proving compliance with an ever-increasing number of regulations.
Far from being a small improvement in an obscure niche of limited interest, this advance could turn out to be the enabling step for a broad range of technology changes.