CS8803ISA Course Reading Summaries Paper #: 5.2 - 16 Title: Effects of power conversion, wireless coverage and cooperation on data dissemination among mobile devices (1) Problems The system addresses the problem of providing data sharing among mobile, wireless devices. The idea is to increase data availability to mobile users in a metro area, where they have only intermittent connectivity to the Internet. Since mobile devices don't get have consistent, cheap access to the Internet, we need ways to share data that don't require such access. The kind of data that people need tends to be localized; i.e., folks around a movie theater might all need to access a web page to buy tickets. The system exploits such localization and tries to keep up-to-date much-needed information available, even without Internet availability. (2) New Idea and Strengths The system does what it can, when it can. It tries to guess what users might need, and get it ahead of time (prefetch). Also of course, when a user asks for something, the system makes specific demands to get that data (if it hasn't been prefetched). The system is effective because it uses a variety of methods to increase data availability. Nodes share the data they have; forward queries and responses to other nodes; and cache popular data objects. It turns out that when the system runs in peer-to-peer mode (like Gnutella, where hosts cooperate with each other), data gets distributed faster among the hosts in the system. The system also has good ways to preserve battery power. When the battery is weaker, it can use less ways to get data, for example, instead of periodically sending out a data query for something it needs (or thinks it will need), it can wait passively for an "advertisement" from a host, i.e., a description of information and/or applications it has. Upon receipt of the advertisement, it can send the query to that host, knowing that it will be satisfied. It's neat that the system can be configured to automatically re-fetch items that have gotten out-of-date, so when the user needs such an item he can access an up-to-date version. (3) Weaknesses and Extensions The paper uses simulations and a rudimentary mathematical analysis to show that the system works. To really know it works, we need to see an experiment that uses real devices, lots of them, in some realistic setting (like an airport or a busy downdown area). The authors propose a neat extension, a mechanism to indicate the appropriate interaction (P-P, S-C, etc.) based on parameters such as data availability prediction, battery level, and proximity of other 7DS devices/applications. They planned to use the application with a tou guide and academic news notification system. There are significant user privacy and security issues to deal with. Any system like this (in its infancy) may not seem very useful and convenient, but this is a start and I think it's promising. However, if we soon get to a point where wireless Internet coverage is more available, this research may be superceeded by plain Internet access by mobile devices.