Quality of information functions for networked applications

Abstract

In this paper we propose a framework for providing quality of information (QoI) aware networking. QoI quantifies how useful information is for an application. Its value is comprised of both intrinsic and contextual attributes related to the information. Intrinsic attributes include freshness and accuracy of the information. Contextual metrics include completeness and timeliness. To design network control algorithms, such as resource control algorithms, attributes of data must be mapped to QoI. Then, network algorithms may deliver the data in such a way that the ultimate information derived from the data is of sufficient quality for its purpose. We propose our QoI framework, and present the concept of QoI functions that capture tradeoffs between different attributes of QoI. We use optical character recognition (OCR) as an exemplar image processing application. We focus on two attributes of QoI for the OCR application: accuracy and timeliness. We show how network controls and data processing, such as error recovery and compression, operating on data, impact the QoI delivered to a recipient. We then show how reductions in the required QoI may have a drastic impact on the amount of network resources required to support a QoI-aware transaction.