By Marco Naveda, Dmitri Fedorov & Raghu Ranganathan, Ciena
The networking software industry is experiencing an accelerating technology shift from centralized data center application delivery models to a distributed edge computing paradigm. In this new computing paradigm, cloud infrastructure and services are delivered from multiple distinct and geographically distributed locations in closer proximity to the end user or data source. The drivers for this shift are the emergence of advanced Enterprise use cases & applications, network infrastructure convergence and operator digital transformation initiatives. These applications require deterministic latency and ultra-fast response times, distributed processing of large volumes of data near the source and flexible scalability across network and computing ecosystems. Included in this transition, the convergence between wireless & wireline, combined with increased hub-site network capacity in cable Distributed Access Architectures (DAA) and the drive to centralized mobile Radio Access Networks (cRAN), are making distributed edge computing more feasible for operators.
This computing paradigm is enabled by cloud-native principles and application containerization & orchestration technologies that promise new operational efficiencies and agility to develop innovative services. We see this trend across Enterprises and network operators that want to accelerate software delivery while maintaining a consistent quality of experience from applications & data to devices and end-users alike, independent of location. 5G and high-speed fixed broadband connectivity services are acting as catalysts to enhance the underlying network performance and agility using a cloud-native services-based architectures. This in turn poses a challenge of operationalizing an applications-first approach in the operator’s network to facilitate open, modular, and portable multi-vendor networking software across their distributed edge compute infrastructure platforms.
Advanced cloud-based management & orchestration systems, however, were not designed for distributed edge computing. These systems were optimized for very large compute environments where server clusters are co-located and mesh inter-connected by an over-provisioned data center fabric. In edge centric architectures, compute & network nodes are geographically distributed and inter-connected by multi-layer access and aggregation networks with varying degrees of capacity, latency, and flexibility.
This creates the opportunity to adapt and optimize the use of containerization and cloud orchestration technologies to meet the needs of embedded real-time network functions and edge business applications.
This paper will delve into emerging edge computing infrastructure architectures, available open-source software to enable distributed edge computing, and key technical considerations to accelerate adoption of new software-based operational methods. This paper evaluates the use of open-source projects from the Linux Foundation Edge (LF Edge) and Cloud Native Computing Foundation (CNCF), such as Akraino Edge Stack and Kubernetes, to build and operationalize distributed edge computing networks. This paper will also explore how hyperscale cloud platforms are addressing this technical challenge and how cable MSOs can leverage a rich technology ecosystem to architect open edge technology systems, implement software defined network operations, and monetize new edge-based business services.