Cloud-native orchestration and automation for disaggregated networks
(English) Transport networks are segmented by different technologies and domains, what makes difficult automated end-to-end operations and making direct human intervention needed to deploy connectivity services. On top of that, datacenter operators are achieving efficiency and cost reduction by deco...
| Autor: | |
|---|---|
| Tipo de recurso: | tesis doctoral |
| Estado: | Versión publicada |
| Fecha de publicación: | 2023 |
| País: | España |
| Institución: | CBUC, CESCA |
| Repositorio: | TDR. Tesis Doctorales en Red |
| OAI Identifier: | oai:www.tdx.cat:10803/692198 |
| Acceso en línea: | http://hdl.handle.net/10803/692198 https://dx.doi.org/10.5821/dissertation-2117-414775 |
| Access Level: | acceso abierto |
| Palabra clave: | Àrees temàtiques de la UPC::Enginyeria de la telecomunicació 621.3 |
| Sumario: | (English) Transport networks are segmented by different technologies and domains, what makes difficult automated end-to-end operations and making direct human intervention needed to deploy connectivity services. On top of that, datacenter operators are achieving efficiency and cost reduction by decoupling the software from the hardware, using white-box hardware with open Application Programming Interfaces (APIs), enabling more customization and vendor competition. Still, there are no clear standard data models that describe the device capabilities, attributes, operations and notifications, as well as efficient transport protocols that provides primitives to view and manipulate the data. Moreover, the growth of demanded by users, threaten to exhaust the capacity of Wavelength Division Multiplexing (WDM) optical networks. Extending the spectrum of use can be a short term solution, but in the long term, Spatial Division Multiplexing (SDM) needs to be deployed to keep up with capacity demands. The first objective is to provide network control to multi-layer multi-domain disaggregated packet optical networks. A control architecture for partially disaggregated optical networks is presented, based on a Software Defined Networking (SDN) controller that is able to controll them. It uses an Open Line System (OLS) controller to manage the optical line, using standard interfaces, and open transponders. To alleviate part of this complexity, an architecture for the orchestration of multi-domain multi-layer, packet and optical networks is also presented. The main idea is orchestrating the packet and optical domains using an agnostic centralized orchestrator that is only responsible for domain selection of multi-domain paths, wrapping the underlying domains with SDN controllers with standardiet interfaces. The deployment of SDM links between overloaded WDM nodes is proposed to alleviate links that are reaching full capacity. The presented target architecture consists of OLS controlled domains for WDM and SDM and open transponders following the disaggregated approach. These architectures are demonstrated in real optical testbeds. The second objective is to present an architecture than can keep up with the centralization and the complexity that the SDN approach comes with. A cloud-native architecture for SDN controllers that is able to cope with the complexity and size of transport networks is presented, based on the micro-services software architecture, that allows for self-healing infrastructure, independent auto scaling of micro-services and autonomous deployment and development of each module. Then, the same micro-service architecture is updated with a solution to scale the database, replicating the micro-service. Data coherence must be kept at all times, and micro-services must be aligned so there is no data corruption or misalignment. This is done using NewSQL databases, included in the architecture. The final objective deals with an architecture that joins the data gathering, training of models, and execution, needed to feed data to support multiple operations in the zero-touch management approach. A telemetry-enabled SDN controller architecture is shown, using novel, open, low-latency, and low-overhead telemetry protocols, with a demonstration of it's scalability. The telemetry system is used to provide the deployment of SDM links between WDM nodes. The proposed architecture also allows dynamic provisioning and reconfiguration of the VNT in case of failure or changes in the requested capacity. The SDN centralization allows improved data gathering from all the segments it controls. This way, machine-learning algorithms can benefit from it, getting access to easy data to train their models. An architecture that joins the data gathering, training of the models, and execution, capable of improving multiple operations is presented and showcased with a Quality of Transmission (QoT) predictor using machine-learning. |
|---|