Memory-enabled Autonomic Resilient Networking

Abstract

It is becoming increasingly challenging for the contemporary Internet to satisfy the explosion in networking services and applications with their widely diverse and dynamic QoS requirements and resource demands. For smarter networking management and operations, we hypothesize that next generation networks should be endowed with "memory"-like (i.e., functionally similar to human memory) capabilities and should explicitly manage their resources and services. In this paper, we introduce SmartNet, a new memory-enabled autonomic network architecture that is scalable, resilient and adaptable at runtime. SmartNet's memory, we term NetMem, manages data related to multiple networking concerns, namely application, communication, and resource concerns. SmartNet's NetMem maintains "knowledge"-the fourth networking concern in our networking model-in a distributed knowledge plane that can be rapidly accessed in a simple and systematic way by SmartCells (SmartNet's basic building block) and the services they implement. Each SmartCell has intrinsic monitoring, control and analysis capabilities and can acquire resources on demand. For scalability and to facilitate inter-communications and services' support between SmartCells, regions, termed CellDomains, may be dynamically formed. CellDomains would incorporate favored services or resources. Our simulation results demonstrate an improvement in networking operations with our memory-enabled network management.