Network monitoring – the ‘only way’ to know what you’re really getting from private 5G
This article is taken from a new Enterprise IoT Insights report on Industrial 5G SLAs. The full article, including additional sections, along with the rest of the report – with commentary from the likes of Accenture, Appledore Research, Nokia, Lufthansa Industry Solutions, Vodafone and VoltDB – is available to download here.
Perhaps the best statistic about the rush of service-level agreements (SLAs) that will go to support industrial-grade 5G network deployments in the coming years is from Canadian test and measurement company EXFO, combining forecasts for a couple of different sources. Half of all 5G revenues, it says, will be subject to SLAs, compared to practically zero for 4G-LTE deployments today.
That is a direct quote from Guillaume Briand, product line manager at EXFO, speaking on the Enterprise IoT Insights webinar session that attends this report. The figure is an extrapolation from a 2020-2025 revenue forecast by analyst firm Omdia that says enterprises will account for 70 percent of 5G revenues in the period, and a GSMA Intelligence survey from 2020 that says 70 percent of new 5G applications will require SLAs for quality of service (“and experience”).
The maths (70 percent of 70 percent) says a whole bunch of critical SLA-writing between telecoms and industry will be set down in the coming years. Interestingly, to put the revenue forecast in context, Omdia predicts a $5 billion boom-time for the widening vendor community just from the sale of private cellular to enterprises, topped up by a ‘ripple effect’ of associated spend on sundry sensing and sense-making (IoT and AI) gadgetry.For its part, EXFO is most interested in supplying test, measurement, and assurance tools to the operator community, which is generally going at the Industry 4.0 market with a two-speed offer: with slices of its public networks; and with various ‘private’ edge networking arrangements, which will commonly use a shared core infrastructure and make use of ‘public’ spectrum. But there are exceptions, and carriers are increasingly looking to extend their management expertise into private standalone 5G setups, as well.
Either way, and every way, SLAs will go against these networks. “Enterprises and industrial application providers will not buy 5G SLAs without performance guarantees,” comments Briand. EXFO, with over 95 percent of service providers on its books, and over 2,000 staff worldwide, reckons it is best positioned to provide service providers with the tools to monitor enterprise customers’ 5G networks, and play “referee” between the sides.
A perfect storm is brewing, says Briand, buffeted on one side by the supply of new network infrastructure and automation tools, and rising analytics capabilities, and on the other by demand from industrial markets. “It is evolving quite rapidly, with the global rollout of 5G, and with the availability of AI and automation in the networks. And Covid-19 has changed the demand for these technologies, as well, just because all the home-working is putting strain on cellular infrastructure, especially,” he says.
“But frankly, with 5G, it is not about subscribers anymore – it is about machines and robots in enterprises. These companies need to be able to rely completely on this new infrastructure. Network providers have to be able to commit themselves to customers with SLAs. The only way they can do this is by monitoring their 5G networks, and only companies like EXFO have the tools for them to measure exactly how these networks are working – to confirm for enterprises they are getting what they paid for.”
In theory, EXFO provides the full “multi-vendor, multi-tech” suite for any network owner looking for a best-of-breed 5G network, with test and measurement solutions to kick the tyres before hitting the road, and assurance and correction tools to make sure the engine is ticking over and the ground is being covered. Briand is responsible for EXFO’s analytics-based assurance solutions, which provide real-time insights for network and service operations centres.
The step-up from best-effort LTE networks, mostly geared one way for consumers, to deterministic 5G networks, for myriad SLA-bound industrial purposes, goes forwards and upwards, suggests Briand. With industrial 5G, key performance indicators (KPIs) are multiplied in both volume and magnitude, and monitoring tools are extended across domains at the same time. “Tolerance is less [forgiving] and more varied with machines,” he says.
“For service providers, wireline connectivity requires SLAs to guarantee availability, bandwidth, latency, performance, quality of experience – all of these KPIs, which we all know. But 5G will be different, and the SLAs will not be so simple. Enterprises will require complete transparency into 5G network and edge compute performance, to get information about mobility, and information per-user and per-device, in real time. Beyond this, this information should be accessible in real time [to other systems] through other APIs.”
The difference with consumer-grade networks, covering most management functions on LTE, is that coverage is the king metric, as a rule; applications may put different strains on the network, but the only headache until now, simplistically, has been whether a video stream drops out on the commute home. With industrial-grade 5G, the range of performance parameters appears to be never-ending, potentially – with a stretch-total arguably equivalent to the number of applications in all of the enterprises, in the world.
“It’s the classic case that one size doesn’t fit all. SLAs have to be individually defined in each enterprise and dynamically fixed and adapted for different classes of machines. Industrial 5G networks should be anticipated. Because it is not just about ensuring massive throughput, but about allocating for each use case – just like the network demands are different for messaging and video streaming on smartphones,” comments Briand.
“Industrial applications make different demands for latency and throughput, and other measures. But the stakes are higher. Imagine you have a factory full of robots, and the connection goes down, and you’re forced to send staff onto the floor to heal the network and reconnect the robots. That’s a big issue for an enterprise.”
Network slicing, whether of public airwaves in hybrid 5G networks or privately-licensed spectrum in standalone edge setups, will allow performance specifications to be allocated either to certain ‘verticals’ or to certain use cases. “There will be slices for gaming or for transportation, and of course for industry. A factory owner or plant manager will be able to put robots and machines on 5G slices. This functionality will be critical for industry, and will provide a means to ensure latency is low and reliability is high.”
It will be critical for network operators, as well, which imagine themselves surging out of a kind of enforced convalescence, after failing to capitalise on generation-after-generation of cellular technology, offering increasing consumer benefits for the same returns. “The ambition for operators is to monetise this dynamic quality-of-service.”
Briand explains: “Some enterprises will pay a premium to get higher quality, lower latency, higher reliability – to connect their critical machinery. Others just want to store inventory on site, at the edge, and don’t need an ultra-reliable network; just a geolocation function to know where things are. You look at the warehousing, say, which wants to connect both the parcels and the robots carrying them – they only need the premium slice for the robots, so don’t collapse or collide.”
But the major challenge to underwrite 5G management is a straight data processing problem, about the volume and velocity of the discipline. Managing the great flood from 5G-connected industrial sensors is labour-intensive and time-pressured. Network assurance tools are required to cleanse and sanctify data streams so they are channelled back into the enterprise to bring new life. There can be no slow-down, says Briand, to pick-out and strike-out networking anomalies, even with such heavy workloads.
He explains: “It is critical that the big data analytics running the AI-driven string processing, required to handle the volume of data that will come with 5G, is not delayed.” All this AI donkey-work must function for both network managers, as well, he says again, so the enterprise customer and the service provider can collaborate on any alarms raised in the system. “It has to provide full-stack visibility at the same time to identify where the fault originates.”
EXFO’s Nova-branded family of ‘network intelligence’ solutions collects and shares 5G events, and maps them to the network topology, jettisoning redundant network data to keep only the ‘good bits’. “This information will be essential for operators, not just for SLA reporting, but also to enable the network automation and orchestration to create dynamic 5G services,” comments Briand.
“The enterprise will be able to specify their own requirements. They will have access to [network functions] on applications and platforms. Network operators are not just running a pipe, anymore, linking machines together, and to applications on top. Their role will also be to provide the right access at the right time, with the right quality – not with more and not with less; but what the customer specifies for the use case.”
He returns, again, to the back-end analytics, to enable this combination of slice provisioning and assurance. “We are not collecting and processing all the events, just the ones we need. And with all this enhanced data, SLA performance can be monitored, and anomalies can be detected and predicted, and the constant impact can be assessed – and issues can be identified before they occur. But the point is it is not just about collecting the data, but understanding what to do with it, and how to extract value from it.”
The point about “full-stack visibility” is an important one. As the rest of this report makes clear, enterprises want control of their systems, to a greater or lesser degree; the role of operators in this is not yet defined, and they face competition from a familiar stable of specialist system integrators, which have served enterprises with most other aspects of their bespoke networking estates.
Briand says industrial 5G management is a collaborative endeavour, to an extent; carriers must be able to share network performance dashboards with enterprises, whether they have been engaged in the discipline to provide first, second, or third line support. The software will need to dovetail, as well, with fault management and performance management systems, and other service inventories. “Because, tomorrow, you won’t call the customer care when issues arise in the network; you will access a self-care application.”
He explains: “Enterprise customers are not just paying for connectivity; they want visibility into their compute performance and availability. The challenge comes from the fact many of these services will run in encrypted slices – and that kind of full-stack visibility is something most traditional networks are blind to. And critical services and closed-loop industrial applications also come with legal and human safety consequences, which mandate precise actionable insights within milliseconds.”
He steps back, and zooms out. Mobile networks have never been so valuable before, he says, because of the value of the cargo they are being asked to carry; but the challenge, at the same time, to keep the burden secure and punctual, has never been so complex. “5G brings a paradox,” he begins.
“The telecoms industry has never had so much valuable data before. But to succeed, and to provide these industrial 5G SLAs without storing and processing everything, AI has a key role to play – in real-time pattern recognition and predictive analysis – to analyse the context of issues arising in the network through through a process of correlation, in order to automatically find out the root cause and to generate SLA-based 5G rules.”