In this paper, Villegas et al. tackle the problem of federating different cloud infrastructures (with the same service interfaces) by leveraging a layered service model. The layers are those defined by the National Institute of Standards and Technology (NIST): Internet as a service (IaaS), platform as a service (PaaS), and software as a service (SaaS). The authors provide a real-world implementation of a central processing unit (CPU)-intensive application, WRF, for weather forecasts. They decouple the layers to achieve the goal of outsourcing different components to cloud providers. Their wish is to have cooperating providers of cloud services, so that when a customer submits a request to one cloud, it can be fulfilled by another as appropriate.
The ultimate goal of the authors’ research is very challenging. As they point out, standard interfaces are lacking. However, process algebras, like the pi-calculus, are already able to describe such interfaces, and a more expressive algebra (that is, one including negotiation and handshaking between other resources like libraries and licenses) can be built out of initiatives like the OASIS topology and orchestration specification for cloud applications (TOSCA). The semantics aspects of the execution in different layers are challenging.
The topic is worthy of attention. Similar challenges have been tackled and partially solved, for example, by the worldwide Large Hadron Collider (LHC) computing grid. This paper describes a very good deployment, and would make a good topic for a workshop.