Collaboration out of the box: A framework for enabling large-scale, energy-efficient pervasive computing

The number of mobile phone subscribers is reaching 5 billion. Mobile phones are increasingly used for collaborative applications, such as IM, Facebook, and those for traffic and environmental sensing. As these applications are becoming pervasive and large-scale, much is left to be desired in terms of energy efficiency. These applications are collaborative and must share information among the participating devices. Basically each device needs to notify peer devices of its latest state and receive new updates from their peers. All the communications must happen in a timely manner. To save energy, they often do not communicate with each other directly. Instead, they are configured to communicate via a conceptually centralized server, or the “cloud” in the modern terminology. Indeed, offloading communication from mobile devices to the cloud solves part of the energy efficiency problem. However, it is practically expensive to maintain a live connection between a mobile client and the server, especially when the client is hosted in a Web browser. Most existing systems cannot proactively notify a client of updates and rely on mobile clients periodically polling the cloud for updates that may become available at any time, or not exist at all. As a consequence, a mobile client has to compromise between timeliness of information and energy efficiency. Meanwhile, the “cloud” also becomes an energy hog if all clients poll it for updates blindly and frequently. To fundamentally address this problem, we have to look into the “cloud” to include the network carriers, a critical part that is often missing in many research proposals. Most previous proposals focus on optimizations in the box – how the device hardware and software can be optimized for energy efficiency. They are reasonable for most single-user applications. However, for collaborative applications that are mobile, cloud-based and large-scale, a better strategy would be to collaboratively optimize all the involved parts together, including the server, the network infrastructure, the devices, and the clients. Optimization techniques in-the-box and out-the-box are complementary. Intuitively, the server and the cellular base stations are connected. The server knows about new updates and the base stations are always connected to the mobile devices as is necessary for uninterrupted service. Hence, theoretically, the server can always push updates to a mobile client’s current service station, which in turn can push the updates to the client or suggests the client to fetch the updates from the server.