Network cloud has been gaining considerable momentum in the industry, both by service providers and cloud providers. The benefits it brings are well understood, and include significant cost savings, leading to a more sustainable and profitable business plan for operators; optimal scalability, allowing networks to flexibly and efficiently overcome the never-ending surge in capacity demand; and software-paced innovation, which allows for the rapid introduction and application of new services, resulting in new revenue streams and improved competitiveness.
To achieve these goals goes beyond selecting a basic, disaggregated routing platform. It is necessary to build a network cloud in an efficient and cost-effective way, which demands four main elements: disaggregation, distribution, containerization, and orchestration.
Hardware-software disaggregation
Network disaggregation is not new, going back to when software-defined networking (SDN) and network functions virtualization (NFV) were introduced almost ten years ago. Hardware-software disaggregation – or vertical disaggregation – makes the networking challenge a software-based one. Using commercial, off-the-shelf hardware, such as servers – it helps simplify the network into common building blocks from multiple vendors.
White-box-based disaggregation, however, is a more cost-effective solution, delivering better performance, as the data plane runs on the networking white boxes while the control plane runs on standard x86 servers. This way both can grow independently of each other to a very large scale by adding more white boxes or servers, respectively. Disaggregation better adheres to the operator’s budgetary constraints, adding levels of flexibility and scalability that cannot be achieved in a traditional, monolithic solution.
Distribution across multiple hardware instances
While disaggregation was a big step towards the evolution of our networks, it was still not enough to address the needs of the market. Running a software-based network function (NF) over a single white box is limited in terms of capacity, interfaces, TCAM size and other parameters – by the capabilities of that one white box. To scale the solution optimally, without the need for multiple sizes and types of white boxes, it is necessary to distribute this NF service instance (SI) across multiple boxes.
As such, a smart abstraction layer is necessary to take a cluster of white boxes and distribute all the resources of those boxes into a single pool of resources. This can be achieved with software, while allowing the SI that runs above it to “see” the cluster of white boxes as if it was a single hardware entity. With this distribution, it is possible to use a couple of basic hardware building blocks to build a router of any scale by simply multiplying those building blocks into a cluster. This cluster is handled as a single hardware instance, with all the resources in the boxes accessible by any software that runs on top of the abstraction layer.
Containerization of multiple software instances
Containerization takes virtualization to a new level. While virtualization is part of a cloud-native network, with containerization it is possible to optimize cloud technology to bring further benefits to network operators.
While running a single network function over a cluster of white boxes is a significant enhancement, it is not necessarily the optimal use for those valuable hardware resources. To better utilize those boxes, multiple services must run over the same infrastructure, leveraging cross-service synergies to fully utilize the hardware resources.
With a distributed hardware pool of resources, each service must be run as a service instance in a container. This well-established concept turns the disaggregated distributed network into a cloud-native network solution – or, in other words, a network cloud. The network cloud allows multiple network functions, from one or multiple software vendors, to run over a cluster of commercial off-the-shelf white boxes, from one or multiple hardware vendors. For operators, each network function appears as a separate entity, with dedicated hardware resources and complete separation from other service instances running on the same cluster.
With networking services running in independent containers, a cloud-native software allows services, like core and edge routing, security and others, to scale independently from one another. This way, large-scale, co-located, software-based networks (such as core, edge, aggregation and peering) can share a single physical cluster, gaining incredible operational and cost efficiencies.
Full orchestration
The last step in the evolution towards the right network cloud addresses the planning, management, and maintenance of the software-based, cloud-native network.
A network cloud solution must include an effective orchestration layer that provides multiple views used by multiple network-operation and infrastructure-operation functions. With one view, it is possible to simplify the handling of the hardware and software infrastructure instances by automating the bring-into-service, upgrading, and troubleshooting tasks of this layer. Another view of the orchestration system provides a monolithic-like view of each network function service instance. As such, even though the service instance runs as a container on top of multiple distributed software and hardware layers, the network function view presents this SI, with its allocated resources, in a simple-to-manage manner.
Full orchestration automates the deployment, scaling and management of the network. Orchestration delivers advanced automation, enhanced visibility and smart insights – from service instances to the entire network. It also automates resource lifecycle management across nodes, clusters, and the entire network, while orchestrating hosted networking or third-party services anywhere, any time and at any scale.
Orchestration eliminates operational complexity by managing the entire network as an integrated single-vendor network, accelerating troubleshooting, increasing availability, and ensuring optimized network availability and performance with granular visibility and real-time insights.
The success of a network cloud
These four elements are not a ‘nice-to-have’, but rather a must to guarantee the evolution of our networks, and deliver lower cost, optimal scaling and ease of innovation.
The first two elements, disaggregation and distribution, result in a distributed disaggregated chassis (DDC) architecture that enables lowest-cost infrastructure and optimal scaling built upon a cluster of white boxes that handles the distribution of software across this cluster and the abstraction of the hardware resources into a shared pool. This cloud-native network relies on lower-cost hardware, translating into the same building blocks for any use case, and optimizing the use of resources, with shared infrastructure and fewer hardware resources.
Containerization allows for optimal scaling and software-paced innovation as it enables the introduction and addition of new services and network functions by software only, over existing hardware infrastructure, with multiple service instances running simultaneously. Last, the fourth element, orchestration, is the ultimate ‘cherry on top’ – the one that simplifies the overall evolution, management and maintenance of today’s network cloud.
