A framework for distributed processing on an offline cell phone network

In the recent years, mobile cloud computing became a vivid topic. On the one hand, the resources of cell phones are rapidly growing, which allow them to solve more and more computationally demanding tasks. On the other hand, cell phones are indeed portable: basically, we can assume that all human beings will be equipped by one. Thus, a scenario aiming to organize these devices into a local supercomputer to further improve computational performance is completely realistic. This paper aims to introduce a framework, which is able to control and schedule distributed processing by cell phones and tablets to solve complex tasks, when an Internet connection is not available. The communication protocol is based on Extensible Messaging and Presence Protocol (XMPP) to take advantage of its platform independence. We investigate whether XMPP is suitable for such mobile applications, and show how our general framework can form a basis to solve specific tasks by a distributed cell phone network. We introduce some measures based on the available resources of the individual devices which let the scheduler to assign the subtasks accordingly. For demonstration, we consider medical image analysis problems, where retinal/dermatoscopy images acquired by mobile phones are processed by a local mobile network.