Network 'memory' system for enhanced network services

The advent of the Internet of things will bring to bear an explosion in the number of interconnected heterogeneous objects as well as the diverse resources and services they may offer. A fundamental goal is to ensure the availability of resources and services to communicating objects ubiquitously, resiliently, on-demand and at low cost while satisfying users' QoS requirements. We hypothesize that to achieve this goal; there is a need to build capabilities for smarter networking to harvest the currently elusive rich semantics that emerge in interactions. In this paper, we propose the concept and primary architecture of a network 'memory' (or NetMem) to support smarter data-driven network operations as a foundational component of next generation networks. Guided by the fact that networking activities exhibit spatiotemporal data patterns, we design NetMem to mimic functionalities of the human memory. NetMem provides capabilities for semantics management through integrating data virtualization, cloud-like scalable storage, associative rule learning and predictive analytics. NetMem provides associative access to data patterns and relevant derived semantics to enable enhancements in decision making, QoS guarantees and utilization of resources, early anomaly detection, and more accurate behavior prediction. We evaluate NetMem using simulation. Preliminary results demonstrate the positive impact of NetMem on various network management operations.