The extension of common cloud platforms from center to edge will be a key theme of Mobile World Congress this year. This may not seem like a mainstream issue for a mobile show, but it is a crucial one for operators. Edge computing and network slicing will be essential enablers of the 5G business model, and particularly the goal of extending 5G services beyond mobile broadband and into many enterprise sectors with specialized needs – needs which will change as they start to look towards mass-scale Internet of Things (IoT) applications.
Advanced operators are planning to deploy cloud platforms, or work with those of partners, in order to support their own virtualized networks and new operational processes, and to deliver services to clients, especially when they move into telco cloud, edge computing and slicing environments. Some other operators will confine their cloud activities mainly to their own internal platforms and processes, and steer clear of moving well beyond connectivity in terms of their services. But even these will need to understand how to optimize their economics by adopting common platforms from center to edge – even if those are being deployed by third parties.
So the benefits of 5G will be maximized when the networks are aligned very closely to a dynamic, programmable and unified cloud-to-edge platform. The enablers of this will be prominent in Barcelona, from chip to fabric to VNF (virtual network function) layers.
Kaloom’s 5G-focused edge fabric highlights telco challenges in this market
A couple of months after raising $10m in Series A1 funding, Kaloom – a start-up specializing in programmable data center fabric – has launched a product for the edge. Called Cloud Edge Fabric (CEF), the system is heavily geared to 5G, promising to help fulfil several key promises of that technology, including very low latency and network slicing. We are highlighting this announcement, before diving into other aspects of edge evolution in the 5G era, because it epitomizes many of the most important trends, and also the risks that operators face of losing the initiative, and their place in the value chain, to those with a stronger data center and digital infrastructure heritage.
Canada-based Kaloom describes CEF as a fully automated data center network fabric, which integrates virtual and cloud network functions (VNF/CNF), with native support for network slicing, and with embedded 5G user plane function. Like SDF, the CEF runs on Red Hat’s OpenShift Container Platform and Red Hat OpenStack Platform and can run network functions in virtual machines as well as hosting them in containers.
“We believe that automation, unified VNF/CNF, programmability and network slicing will disrupt the way service providers deploy and manage cloud edge data centers,” said CEO Laurent Marchand.
Kaloom said it is currently in trials with several (unnamed) tier 1 telcos and aims to have the product commercially available later this year. Thomas Eklund, Kaloom’s VP of marketing and strategy, said: “We have done a very good 5G slicing solution. A key feature in our software defined fabric is that you can virtualize the data center.”
The CEF is effectively a shrunken version of its inaugural product, SDF (Software Defined Fabric), but that was targeted at large data centers. The two platforms share the same key capabilities however – programmability; automation; integration of different virtual components such as virtual routers, switches and gateways; data plane offload; and a basis in white box hardware, centered on the Barefoot chip architecture.
The CEF is heavily geared to 5G roll-outs. One of its key capabilities is to support the 3GPP 5G standards for User Plane Function (UPF), a central element in the 5G systems architecture. Kaloom’s implementation of UPF allows packet core applications to be offloaded to the edge and scaled up and down from there, says the vendor. Because CEF is programmable using the standard P4 language, customers can add their own functionality to UPF and link to third party services.
“One of the things we wanted to do is make sure we’ll be able to host additional network functions as opposed to just the network fabric,” said Kaloom CTO Suresh Krishnan. “That’s why we do the UPF. We can interoperate with any other vendor — Ericsson, Huawei, Samsung, and of their 5G entities.”
Common fabric and software architecture from central cloud to the far edge is important to the economics of the distributed cloud, and therefore its likelihood of getting deployed at scale, in order to support applications that will often, also, be enabled by 5G. The edge cloud and 5G share the potential to be extremely distributed – eventually, with 5G small cells and tiny edge nodes integrated in almost any location.
They also share the same differentiators from their older relatives, centralized cloud and 4G. These include very low latency, automation, and the ability to support a high level of control over security, management and orchestration by keeping the whole network on-premise – a self-contained virtualized RAN and packet core, plus an edge node, linked back to the main core and the central cloud by fiber or 5G.
But who is defining the rules for this era in which cloud computing and 5G will become so tightly intertwined at the edge? Much of the innovation is coming from the webscale and data center players, which gives them the opportunity to increase their penetration of telco accounts – but may disadvantage the incumbent network vendors, while forcing operators not just to deploy new kinds of platforms, but to work with new suppliers from a different world.
An example is seen in the way Kaloom’s fabric natively supports network slicing. Slicing is the technology which, in many respects, will make sense of the 5G model once it extends beyond conventional MNO services and into mass-scale industrial and IoT applications In slicing, each virtual segment of the network can be provisioned and optimized for a specific use case, enterprise or user group – statically at first, but eventually on a dynamic, on-demand basis.
In CEF, there is full isolation of each slice or tenant, right down to hardware level, to maximize security, says the supplier. To date, Kaloom claims to be the only vendor doing this hardware-level slicing, though it expects Cisco and others to follow soon.
Hardware-level slicing enables the slice to extend across many data centers and to the edge, and to grow and shrink according to usage. While many operators are still talking about slicing in terms of creating just three or four segments to support very broad use cases and sectors (a generic low latency slice, or automotive slice, for instance), many cloud platform providers are already thinking in far more granular terms.
It’s another example of some operators being well behind the curve in their thinking. In slicing as in edge computing, many operators are still focusing on a few well-understood services. So they consider edge nodes mainly to enhance existing applications like video, which can be delivered from a few locations aligned to their sites; while the cloud market is thinking about huge numbers of edge locations, often indoors, to support a wide variety of enterprises and applications. Likewise, the enterprise and webscale suppliers are thinking in a far more granular way about slicing, and particularly that the 5G cloud and edge will have to support large numbers of diverse, mainly enterprise use cases.
So the launch highlights the challenge to traditional telcos and their suppliers, in the deepening intersection between 5G and the rising interest in edge computing. This is an immature market in which there is still a chance for start-ups like Kaloom to make an impact, despite the activities of the IT giants like HPE and Cisco; and many of the most interesting solutions which are emerging to support 5G-related concepts like slicing are coming from a cloud or data center heritage, not from the telco market at all.