SPEAKER: Antonio Manzalini, TIM
Networks are undergoing an unprecedented techno-economic transformation driven by the ever growing deployment of IT equipment within communication infrastructures, with the prime aim to realize most legacy functions in software. This trend will ultimately deploy a large amount of computing and storage resources into the network, building a vast programmable execution environment that open the opportunities for far more services beyond mere voice/data transport (5G networks). Telecoms are carefully looking at this evolution, because of the higher revenue margins for value-added services over their traditional business model.
The key driver in this evolution is the definition of some kind of middleware that is able to run heterogeneous services on heterogeneous and distributed hardware, in a transparent, seamless and automated way. This middleware can be designed in a similar way to a modern operating system, offering standardized interfaces to applications and services (similar to system calls and libraries), abstracting the resource diversity (through some abstraction similar to device drivers). The Network Operating System (NOS) should provide a platform for automated operation, bringing together communication infrastructures, IT equipment, terminals. The definition of a NOS entails several on-going initiatives, including the Everything-as-a-Service (XaaS) unification paradigm, service interfacing and templates (e.g., IETF intent framework, OASIS TOSCA/YANG), virtualization environments (e.g., OpenStack, NFV, ONOS), programming and configuration protocols (OpenFlow, Netconf, …), commoditization of Telco central office and disaggregation into edge/fog components (Telecom Infra Project, Open Compute, Mobile Edge Computing, etc.).
The topic has been discussed for several years, and there are many components that could be assembled together in a common architecture (e.g., CORD2). However, such a distributed and heterogeneous environment will be characterized by a higher level of complexity than today, and this will be a challenging issue for proper automated control and management.
Artificial Intelligence (AI) may be the right solution for complexity. AI has been already used successfully for detection of cyber-attacks, and its usage is under investigation in different sectors: car driving, smart ambient, financial and market prediction, genetics and DNA programming. But the challenging question is: could AI be the “end of code”? May we envision a new paradigm, i.e., AI-defined network for making large-infrastructures and big-data “actionable”?
