The facility to undo a collection of changes, reverting to a previous acceptable state, is widely recognised as valuable support for dependability. In this paper, we consider the particular needs of the user of cloud computing resources, who wishes to manage the resources available to them, for example by installing and configuring virtual machine instances. The restricted management interface provided to the outside by a computational cloud prevents the usual rollback techniques: there is no command to save a complex state or configuration, and each action may have non-obvious constraints and side-effects. We propose an approach which is based on an abstract model of the effects of each available operation. Using this model, those forward operations that are truly irrevocable are replaced by alternatives (such as “pseudo-delete”), and an AI planning technique automatically creates an inverse workflow to take the system back to the desired earlier state, using the available operations. We demonstrate the feasibility and scalability of the approach.

Hope to see you at HotDep 2012 in Hollywood, till then, you can download a copy of this paper from our NICTA Research Publication website.

We will also make the tool available for you to download and use shortly, we’d love your feedback – stay tuned.