Nothing illustrates the way that 5G and the edge cloud will complement one another, than HPE’s newly announced alliance with Samsung to combine their respective edge and virtualized RAN technologies.
Both vendors have the opportunity, at the dawn of 5G, to get deeper into the mobile infrastructure market – HPE because mobile RANs and cores will increasingly be virtualized on servers; Samsung because it has, after years of being a minor player in mobile networks, got several early contracts for its 5G gear, and may be hoping to benefit from the uncertainties surrounding the Chinese suppliers (see separate item).
The partners said they will offer a joint edge-to-core vRAN solution based on Samsung’s radio technology and system integration services, and HPE’s latest edge cloud platform, the HPE Edgeline EL8000 Converged Edge System.
Samsung’s head of technology strategy, Wonil Roh, said the solution would target customers needing to deploy vRANs with very low latency, bringing the baseband functionality closer to the user by running it on an edge node. This will help to “drive the change towards a standard-IT based and software-defined network edge,” he said.
HPE also believes edge computing will be used to virtualize smaller LTE networks to fill in gaps in 5G coverage – which will exist for many years before operators deploy the new network on a nationwide basis – while supporting 5G-like capabilities such as low latency response. In another MNO-focused alliance, it is working with integrator Tech Mahindra to deliver solutions based on the Edgeline EL8000, using the Indian firm’s MEC (Multi-access Edge Computing) software.
The upgraded Edgeline platform is targeting 5G migration and specific 5G-oriented use cases like smart cities, artificial intelligence, video analytics and enhanced media delivery, said HPE.
The product is compact and ruggedized, to maximize the variety of locations where it can be deployed, and to support colocation with remote cell sites. It is based on Intel Xeon Scalable Processors, with up to 1.5TB of memory and 16TB of storage. It is a modular system in which various multivendor system components can be mixed and matched, scaled and hot-swapped according to changing usage patterns and applications. The components include Nvidia’s Tesla graphics processors, Intel and Xilinx FPGAs (field programmable gate arrays), and network interface cards from Intel and Mellanox.
There is also a range of options for blades and chassis, so that the system can be flexibly configured and scaled as requirements change. For instance, the edge platform can be reconfigured by an MNO to support more distributed networks of small cells, or to scale up a cell site and its baseband processing capability as traffic and device density rises. One-click provisioning and remote systems management, monitoring and updates are also supported, using the HPE iLO 5 technology and the firm’s new chassis manager software. This helps an MNO manage the edge systems supporting thousands of cell sites, even very remote ones, without needing IT expertise in those locations.
“CSPs have come a long way in virtualizing their networks. As this continues, the distribution of converged communications and compute capacity throughout the network will accelerate lower costs and improve service,” said HPE’s general manager for converged servers, edge and IoT systems, Dr Tom Bradicich.
Last summer HPE said it would spend $4bn over the next four years on the development of edge products and services, with a focus on intelligent services for emerging applications such as high end security, artificial intelligence and machine learning.
Meanwhile, Cisco has had a significant influence on the early edge market by initiating the ‘fog computing’ concept which is now enshrined in the OpenFog framework, the basis of the IEEE’s standard for edge interoperability. Unlike fully edge-focused architectures like ETSI Multi-access Edge Computing, the idea of fog (and its more dispersed cousin, mist computing), is that data should be processed in the most appropriate location, which could be anywhere from the central cloud to an tiny in-home or enterprise node.
With more and more data being generated, especially by IoT devices, in many cases it will be cost-efficient, and reduce latency, to analyze and filter the data locally, and decide whether to discard or store it, or send it to the central data center. But data and analytics will also need to be addressed at many other locations in the chain, depending on the application and the level of integration needed across a wider area. An intelligent view of the best place to address any piece of data or application requires a unified framework from center to extreme edge, and this is the basis of Cisco’s strategy.
However, the company has its own challenges in translating its influence into dollars, and will especially be looking to cloud and edge offerings to improve its performance with service providers. In its most recent quarterly results, while its enterprise and government divisions showed growth, SPs were still in decline.
One issue for Cisco is that cloud service providers are encroaching on its heartland in enterprise data centers, sometimes using edge computing as a Trojan horse – AWS has said that its Outposts on-premise edge platform is partially a way to encourage cautious enterprises to migrate to the public or hybrid cloud, step by step.
However, according to CEO Chuck Robbins, there is still an opportunity in developments like Outposts. Like operators themselves, Cisco will look to its networking expertise to complement the activities of the cloud giants, which are centralized by nature and only just grappling with the notions of extremely distributed, hyperconnected systems. Robbins told the recent Goldman Sachs Technology and Internet Conference that AWS Outposts with Cisco networking was “certainly conceivable”, though of course AWS would “have to want that and we’d have to work with them” for a deal to be done.
He explained: “Each of the cloud providers is trying to build a simple on-ramp for those customers who want to have their applications run in their private data center as they run in Amazon, for example. There’s an opportunity for us to provide technology that fits into that. [AWS CEO] Andy Jassy and I have a really good relationship. We’ve been talking about what can we do together as companies. They realize it’s not a public cloud-only world.”
This has echoes of the alliance AT&T has with AWS, in which the two companies effectively carved up the enterprise account to play to their respective strengths, with the telco handling connectivity, security, device management and many other network-based services, and AWS supporting the cloud platform from center to edge. This helps AWS fight against the trend for companies to use multiple clouds, moving between them flexibly rather than forming an exclusive relationship with one provider.
As Robbins pointed out, Cisco can bring its edge technology and its high end switches to the party – it now has 400GbE switches geared to webscale providers and large operators, and in December it acquired silicon photonics firm Luxtera to enhance that 400GbE portfolio.