SIs in trouble, hyperscalers indifferent – the case for carrier-led private 5G
Note, this article is continued from a previous post, entitled ‘Software AG, StarHub and the blueprint for carrier-led industrial private 5G’, available here.
Bernd Gross, chief technology officer at Software AG, is presenting the case for telcos as the go-to agents for private 5G. He is discussing the ‘run’ phase, after the sale and build, of private 5G, responding to a question about how the market will be carved up in the management of new digital infrastructure, between the enterprises themselves, with their own spectrum licences, and the various parties engaged in the supply of their shiny new networking gear.
Interestingly, he thinks the case for enterprises to take charge of these networks is overstated. It is a modish trend, he implies, part of the industrial discipline to get to grips with novel technology, symptomatic of a broader move to prime the pump for Industry 4.0. It will not last in any serious fashion, he says. He should be listened to, just because his firm, more than most, is crossing-over into industrial markets, mindful of the intrinsic value of ‘domain’ knowhow.
But he wants to step back; the bigger trend in IT, and also in operational technology (OT) on the factory floor, is to outsource compute workloads, at least. Mostly nothing is coming back in-house. Run-phase management of private 5G in Industry 4.0 must be viewed in this context, he says. “Let’s park 5G for a moment. Because there is this massive trend anyway to outsource IT – to make IT more repeatable, from plant to plant,” he explains.
“Which is driving traditional OT towards the cloud, as well. If you think of 5G coming into that trend, it doesn’t make sense for IT departments to manage the 5G network themselves. You have to differentiate between IT on the IT-side and IT on the OT-side; and IT on the OT-side is already being automated and streamlined, and going into the cloud. There is less and less knowledge on site, and more and more normalisation in the cloud – to scale-up business.”
He adds: “5G is an important complement to digitalise these factory scenarios. But that [process] will not be handled, typically, by the factory owners themselves. In 80 percent of cases, perhaps, factory owners will look to outside partners to help build and run these 5G environments.”
We should distinguish between edge placement and edge management. Gross is not suggesting the infrastructure is being centralised; privately-operated 5G in privately-licensed spectrum will be localised, on premises, and distributed through the edge-cloud continuum in German-style campus-netz. Only the run-phase edge-management – data-crunching, insight-making, network analytics – is going off-site, and being automated, as possible.
Indeed, Gross advocates a hybrid architecture, strung across the site edge, network edge, and national cloud, with the balance between shuffled with latency, reliability, and criticality requirements according to a hierarchy of industrial use cases. “We are talking about all of it, everywhere; it is about lifecycle management – about how to operate and make sure these networks run according to prescribed SLAs. The winning architecture is going to be a hybrid one.”
What is clear, he says, is the tech-side battle over grunt-power is finished; the hyper-scalers have won. One way or another, in one place or another, industrial compute workloads are going to the likes of AWS and Azure, and Google to a lesser extent. “The hyperscalers have so much in the way of economies-of-scale. They have already won on the backbone cloud environment,” he explains.
“You have the hyperscalers and telcos at national level, and all this telco investment to create edge-cloud environments for 5G, and then the on-prem edge. But a lot of the telco clouds are built on hyperscaler infrastructure. It is the same with these industrial giants and big automotive companies. The hyperscalers are becoming infrastructure-as-a-service suppliers; everyone else is reverting to running cloud application environments on top.”
Which is where the opportunity is for the telecoms community, in over-the-top management of edge infrastructure. Because the hyperscalers, he reasons, are not much interested in the nitty-gritty of running industrial operations, even as they have forced the trend to outsource technology services in industry. It does not fit their business model, the logic goes; the ground is left instead for sundry service providers and system integrators.
Gross comments: “The hyperscalers have not yet won on the application [suite] and they have not won on the domain knowhow. And, honestly, they are more interested to get workloads onto their clouds. Something like a third of their revenue is profit – because they have a completely automated and standardised offering. To an extent, they are less interested in very bespoke industrial know-how.”
Which gets us back to this question about who manages the new edge infrastructure – if not the enterprise, typically, and not the hyperscaler, actually. Gross presents three alternatives: the telecoms provider, the industrial specialist, and the system integrator (SI). The first brings cellular know-how, the second brings domain knowledge, and the last… well, what is a system integrator, anyway? Gross has them down as IT types, it seems, running data centres.
In this definition, at least, they are out of the game. He says: “When you look at that anyhow-trend towards the cloud, most of the [OT] workloads are going onto hyper-scaler clouds, onto AWS and Azure. They are not going into the SI clouds; it is always one of the big clouds. So I think the SIs are in a difficult position at the moment – sandwiched between the cloud guys and these tech providers coming in [to the OT space].”
Not everyone agrees; SIs are the flavour of the month in telco-land, especially among 5G kit suppliers for new channels to market, especially into enterprises in Germany (see recent announcements from the likes of Nokia, Airspan Networks, Druid Software). But Germany-based Software AG has a different vantage point, with a developing crossover base in both telecoms and industry, and a prominent position in Industrie 4.0 at home.
And in the end, industrial networking spans these twin disciplines, running vertical in the industrial layer and horizontal in the networking layer. Goss comments: “[There is an opportunity for] telcos with a very compact and compelling proposition to combine 5G campus networks with local cloud services – to bring the compute closer to the factory, and to meet the industrial SLA requirements around latency and other things.
“The vision is to create network coverage in the factory, and to move compute into the local edge cloud – as part of their [application-layer] data centre offering. But you need to think about Siemens, Rockwell, Bosch, and all of these with domain knowhow. Because the chemical industry needs process engineering; manufacturing needs engineering. They have a big future in the edge environment – just not in the infrastructure-as-a-service piece.”
The thing a number of these companies, on both sides, have in common is Software AG. The penny drops, of course. The Darmstadt firm’s Cumulocity IoT platform is being used variously by Deutsche Telekom, Telefónica, KPN, A1, NTT, and Telstra, among 20-odd carrier contracts. Siemens’ Mindsphere platform, leveraging hyperscaler cloud infrastructure, offers extended device connectivity with Cumulocity, although Siemens says it is based on its own IP; there is no word from Gross on Bosch, Rockwell, or General Electric, except to illustrate a point.
The company strategy behind the (compelling) executive dialogue is to sell in the rest of its data analytics and IT/OT integration tools to help with unified management of distributed networking architectures: its Apama event processing engine for real-time analytics, its TrendMiner platform for visualising time-series data, its Analytics Platform for drag-and-drop OT analytics, its webMethods integration server for mashing together IT and OT functions.
“The use cases are different, the edge cases are different, but the cloud platform is very similar. A lot of OT use cases, which used to be in factory pyramids in proprietary IT/OT systems, are moving into hybrid architecture, thanks to 5G. Some of them – analytics for robots, for predictive quality – need massive amounts of data. The paintshop case we have with Dürr relies on 100,000 data points every 20 milliseconds. You need 5G for that.”
You also need a “unified application environment”, he says, to manage it all in the same way – from the Dürr case at the far-edge to the low-level sensor at the near-edge. Enterprises, and enterprise network managers, must be able to “develop once and deploy anywhere”; this is what StarHub in Singapore has in place, expanding a Cumulocity deal for standard IoT into something more profound, which makes 5G a function of a grander industrial-change package.
It is what Gross and the team hope the rest of its carrier customers will do, as well. How is it going? Because it sounds like logical step; it almost sounds like a simple sale. “I wish; it’s not, to be honest. Because not every telco is on board with managed 5G campus networks; not all of them are looking to offer industrial 5G edge compute. But if they make those choices, then I think we are in a very preferable position.”
He adds: “But these ideas – that the cloud infrastructure is distributed, that application environment is unified, that the solutions are developed once and deployed anywhere – represent a very important paradigm for the future. The winners in all of this will be the ones that can operate these hybrid clouds, and offer lifecycle management in a professional manner for the whole architecture. And our technology enables that.”