Dept. of computer science
The global Internet as we know it is changing at the edges. From a edge network dominated by the wired Ethernet, the edge is evolving into disparate wireless ecosystems such as cellular, 802.11, Bluetooth, RFID, sensor networks. The density of end-nodes or devices in these networks expected to be an order of magnitude higher than current networks (100 to 1000 nodes/100 cu-ft from few nodes/100 cu-ft). Deployment and use of the networks will be outside the traditional organization/enterprise boundaries (from being deployed in an administrative domain to being deployed in the field, home, car, road, stores etc). Due to the very nature of deployments and application drivers, data exchange within each ecosystem or across ecosystems will occur when parties are in proximity or near-field. Hence, these networks are termed near-field networks. In these networks, non-traditional metrics such as manageability will take precedence over traditional metrics of performance. Managing such disparate dense networks require augmenting the existing protocol stack for providing efficient remote manageability, diagnosability, configurability and trustability.
The density and the nature of the deployment of near-field networks make it ill suited for traditional network management paradigms. A human system administrator for each network will be prohibitively expensive or even impossible. Management actions that can take human RTT need to be replaced by actions that can be accomplished within packet RTTs. Here, support for remote manageability needs to be part of the protocol stack involving not only network elements but also the end devices. The end devices can serve as alternate paths to conveying state information and possible remedial actions .
Another unique aspect of near-field networks is the nature of data exchange. Data to a consumer is of interest only when it is "close" to the producer of data. Data exchange occurs only when the source and destination are in each others orbit or in a small world setting with not much routing. This calls for the design of opportunistic protocols that can decide when communication occurs and by what conduit based on a discovery, trust, exchange cycle. The design tradeoffs involve the desired trust level and the concomitant protocol complexity .
Finally, the end-nodes will be a significant part of the network and as such configurability and diagnosability of end-nodes should be part of network management. To make this possible, the state of the end-nodes including its capabilities should be exposed in a secure fashion to the management agents. From a management perspective, the end-nodes should be treated as first-class objects making remote configurability and diagnosability possible.
Design for manageability will likely emerge as the key research challenge in near-field networks. Density and the nature of deployment of different wireless ecosystems calls for a new framework that augments the existing protocol stack to support manageability. Support for remote configurability and diagnosability will be critical for the success of these networks as hence knowledge of the capabilities of the end-nodes will be important. The primary focus of networking research should be on improving nuanced metrics such as manageability, diagnosability, configurability and trustability rather than the traditional metrics of performance.
 MIAMI: Methods and Infrastructure for the Assurance of Measurement Information Wade Trappe (WINLAB, Rutgers University), Yanyong Zhang (WINLAB, Rutgers University), Badri Nath (Rutgers University), In the second International VLDB workshop on DMSN 2005, Trondheim, Norway, August 2005
 A knowledge plane for the Internet, David D. Clark, Craig Partridge, J. Christopher Ramming, John T. Wroclawski Proceedings of the 2003 conference on Applications, technologies, architectures, and protocols for computer communications 2003, Karlsruhe, Germany August 25 - 29, 2003
 Architecture and Techniques for Diagnosing Faults in IEEE 802.11 Infrastructure Networks. Atul Adya, Paramvir Bahl, Ranveer Chandra, Lili Qiu, In the Proceedings of the MobiCom Conference, Philadelphia, PA, September 2004.