Science applications expect networks to perform in a reliable and predictable manner to enable the distribution of computation and data movement to produce science. New technologies and new paradigms have emerged in the Research and Education (R&E) Networking community to support science applications: Data Transfer Nodes (DTNs), Programmable Data Planes, In-band Network Telemetry (INT), distributed compute and data infrastructures, research and dedicated network testbeds, and new inter-domain federated orchestrators. Network virtualization and programmable networks are two key enablers that facilitate agile, fast and more efficient network infrastructures, as well as service development, deployment and provisioning.
Challenges in the WAN are being addressed by the adoption of orchestration and automation solutions to improve automated control, configuration and operation of network services, to reduce labor-intensive network deployments. Most network orchestrator solutions require science applications to interact with an orchestrator interface to respond to actions, recommendations, and for service instantiations. As a result, these interactions can result in adding undesired complexity to science applications.
The AtlanticWave-SDX project is developing a closed-loop orchestration system with the capability to react to unplanned network conditions and events by adding autonomic networking functions. The Autonomic Network Architecture (ANA) paradigm defines a set of self-managing functions for self-adaptability and context-aware behavior changes in response to the emergent properties in the networking environment, with limited contextual information and local control. Closed-loop orchestration refers to a continuous and repeating cycle of communications between the network infrastructure and software elements, including analytics, policy, and orchestration, to enable self-management capabilities.
A goal of the AtlanticWave-SDX project is to automatically regulate the participating Open Exchange Points (OXPs) to a desired state without human interaction by creating autonomic control loops through the design and implementation of resource provisioning, control, learning, and fault tolerance algorithms in the policy modules in the AtlanticWave-SDX controller and management components. A major contribution of the project is to implement Closed-loop orchestration on a per-OXP level and inter-domain level (across multiple OXP domains).
For the Community Exchange, we propose to present the AtlanticWave-SDX project as a solution to improve reliability and predictability of network services for science applications. Closed-loop orchestration, autonomic network functions, and how network telemetry and new technologies at each OXP are leveraged to develop end-to-end autonomic network functions will also be presented.
AtlanticWave-SDX is a project at Florida International University (FIU), in collaboration with Renaissance Computing Institute (RENCI), Information Sciences Institute, University of Southern California (USC-ISI), Institute of Informatics of the Federal University of Rio Grande do Sul (UFRGS), Brazil’s National Research and Education Network (RNP), and the Research and Education Network of Sao Paulo (rednesp).