Stateless services are easy to scale in the cloud system because of new
replicas of these services can be created at any time and
they operate independently of other instances.
In contrast, stateful services, such as a database system,
can take minutes up to hours depend on the need to
respond each users of the service.
Currently, this problem is solved by resource overprovisioning
in anticipation of demand spikes. FlurryDB
uses virtual machine cloning to improve resource utilization
by drastically reducing the latency required to add a new
replica. We also show that FlurryDB is capable of handling
updates to resources in a fashion that preserves consistency
across the cloning boundary.