Research database

With 5G and beyond, network architectures are facing unprecedented changes. Cloud resources are being pushed at the edge of the network to meet latency requirements of critical applications and to keep sensitive data in trusted locations. Moreover, network technology is radically changing while embracing a programmability approach that enables very efficient traffic manipulation and computing tasks directly in HW pipelines (e.g., P4) once used only for switching, while SW approaches (e.g., eBPF/XDP, DPDK, smartNICs) become increasingly important because of their close vicinity with end-user (cloud-native) services. However, current network programmability primitives target mainly network experts, with a limited capability to accommodate the necessities from other domains (e.g., cloud services). NEWTON will address the above problems by designing and developing a novel framework that will expose network programmability primitives to a larger community, such as cloud-native developers and cloud orchestrators. For instance, “cloud native” services ask for new requirements (e.g., seamless scalability, unprecedented agility, geographical replicas, controlled latency between elementary microservices, continuous integration and delivery, and more) that demand massive cooperation from the underlying network platform. Furthermore, given that agility has no value without efficiency, NEWTON pursues an aggressive decomposition of network functions into “micro” (and distributed) elementary components, possibly deployed on multiple target platforms (e.g., SW, HW, smartNICs), running separately and dynamically orchestrated (by the NEWTON platform) to seamlessly provide the expected behavior. As part of the emerging network programmability paradigms, NEWTON includes also novel technologies such as Segment Routing, which provides the ability to code directly into each packet header where the traffic should be sent and how the traffic should be treated, thus representing an ideal complement for its programmable network functions. In NEWTON, users will be able to ask for complex services using either imperative and declarative abstractions (i.e., intents), as well as design their own service by composing the available elementary building blocks. The richness of the underlying infrastructure will also be exposed to developers, with different abstractions available to different categories of developers. The NEWTON versatility in the exposure of abstractions, APIs, tools and platforms is of fundamental importance to foster the adoption of the foreseen platform and stimulate faster technology progress. Finally, NEWTON will also target network developers and experimenters by providing a prototyping and testing platform, enabling the above actors to practice with advanced network programmability concepts and emulate their behavior ahead of any field trial.