The rising of applications with intense requirements in data volumes, storage space and CPU/GPU utilization, such as Machine Learning/Artificial Intelligence (ML/AI) applications, imposes different challenges on the Data Center Network design and operation. When compared to traditional data centers infrastructures, the recent explored disaggregated optical data center concept may bring multiple benefits in terms of optimized usage of the IT and network resources. Nevertheless, at the same time, it also brings some technical challenges. In such context, the paper discusses both data and control architectural solutions for optical disaggregated data centers for ML/AI applications, focusing on their benefits but also on the associated complexities.
Future 5G and beyond services rely on the network slicing concept, in which underlying network elements are split and/or aggregated to compose a synthetic network infrastructure (the slice) to satisfy the requirements of services that will be executed on top. Generally speaking, end-to-end network slices comprise multiple network segments, including optical and data centers networks. Therefore, the provisioning of end-to-end network slices is a challenging task that has to consider the characteristics of the different technologies to satisfactorily map the requirements coming/imposed from the services to be deployed. This requires that offers towards the fulfillment of the services to be supported are properly parametrized, enabling the possibility to translate them into specific slice and network services characteristics to be finally materialized in concrete infrastructure resources. On the other hand, there is a rising trend of quality assurance at all levels to satisfy the requirements of services deployed, requiring the runtime maintenance of quality of service/experience of the deployed slices. Due to the dynamic nature of services, it becomes essential to monitor the associated Key Performance Indicators (KPIs), derive from them current quality levels and implement the necessary mechanisms to steer the behavior of the slices towards the maintenance of optimal quality levels. Given such scenarios, in this paper we present a framework that enables the provisioning and orchestration of network slices in multi-domain/segment optical networks as well as an approach to proactively manage the maintenance of the required slices quality. The presented framework is validated through several experimental results.
Network slicing with Quality of Experience/Quality of Service (QoE/QoS) guarantees is seen as one of the key enablers of future 5G networks. Nevertheless, it poses several challenges in both resource provisioning and management that need to be addressed for the efficient end-to-end service delivery. In particular, network slice deployments considering operation across several domains and network segments, require of inter-domain configurations, continuous monitoring, potential actuations, inter-slice isolation, among other, in order to be provisioned and maintained, looking forward to guaranteeing their assured Key Performance Indicators (KPIs). In such scenario, optical networks are of prime importance, enabling the inter-connectivity between multiple far away segments and Points of Presence (PoPs). In light of this, in this paper we present an architecture design enabling network slice provisioning for 5G service chaining in multi-segment/multi-domain optical network scenarios. The presented design is enriched with a policy-based monitoring and actuation framework able to maintain the desired QoS for the provisioned end-to-end (E2E) network slice. We experimentally validated the proposal against real slice deployments and traffic generation, providing a proof of concept for the presented architecture, with special emphasis in the demonstration of the actuation framework as a key element for quality guarantees.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.