5G has far broader ambitions than being another generation of mobile technology that primarily services consumers with voice and data services. As Spirent Communications’ Stephen Douglas, head of market strategy, put it at the Mobile Edge Forum virtual event, 5G “has these ambitions to become an enabler for all types of economic sectors, including our manufacturing plants and processes, our transportation systems, healthcare, the energy sector, and mobile gaming, to name a few.”
What do all of those sectors have in common? Strict demands for both low latency and reliability in communications, Douglas says. Mobile collaborative gaming, for instance, targets end-to-end latency of 20-25 milliseconds for smooth and responsive play. Time-critical sensing applications in transport and logistics aim for less than 30 ms of end-to-end latency. But, he adds, most cellular networks have latencies that fluctuate between 20 to more than 75 milliseconds. There are real-world consequences to high latency: In immersive digital environments with interactive graphics, high latency can trigger “cyber sickness”. In automotive applications, remote driving reactions and interactions can be impacted and lead to accidents or deaths. In manufacturing, degradations in performance and accuracy could cost millions in lost output and downtime.
Jitter—the variance in latency across the network—comes into play as well, because inconsistent latency can also wreak havoc with user experience, Douglas notes. He cited the example of virtual reality where high jitter may delay one headset visual channel differently than the other—again, potentially resulting in headaches or nausea for the viewer.
“If 5G is going to serve these economic sectors, then it needs a plan on how to deliver reliable low latency,” Douglas says.
And it’s not just human user perception and acceptable tolerances that are in play, Douglas notes. “We also have to realize that not all users are humans,” he says. “While a human may be able to tolerate a reasonable degree of variance or inconsistency and latency, machines cannot. Mission-critical machines … need predictable latency that is tightly bounded by permissible maximums and minimums. A highly inconsistent network with fluctuating latency is not ideal for any type of industry adoption, nor for monetizing it as a guaranteed [service-level agreement].”
So guaranteeing a specific and reliable level of latency is key to providing and monetizing 5G services. But that’s a tough challenge, because latency is impacted by many facets of the network from the air interface and radio to core network processing and distributed locations, the transport stretching among those locations, applications and their own processing—some of which are in network operators’ control, and others which are external to the network. The industry is attempting to reduce both latency and jitter by bringing applications and services to the network edge through multi-access edge computing, Douglas pointed out. An “edge” may be at a peering point or peripheral of a telecom network in a major city, which can reduce latency to around 20 ms; other options for edge placement include transport aggregation points such as central offices, or even further into the RAN, or on-premise at customer locations as part of a private network, which “could potentially unlock latencies even lower due to the extreme localization of processing,” Douglas says.
Both public and private edge offerings are taking shape, with telecom providers often working with cloud partners to facilitate them—which raises new questions about delivering low and consistent latency. For example, can public cloud really guarantee the necessary determinism for SLAs? Are edge locations capable enough or close enough to the users? And will enterprises pay for certainty in latency?
Spirent commissioned a survey earlier this year that asked around 150 North American businesses about their early interest in using MEC. According to that survey, 56% of the companies which participated indicated that they would be willing to pay for an SLA with a guaranteed latency window. “Low latency alone is not enough; it must be consistent and bounded in a window that it never passes outside, as this is what the application will require for determinism,” Douglas said. The good news, he added, is that there was “genuine consensus” around latency windows which businesses wanted: 20-50 ms and 10-20 ms.
The survey, Douglas said, indicated “clear evidence of a sweet spot for low latency services where enterprises would be willing to pay for, but this opportunity may not be achievable or monetizeable until consistently low latency can be proven and guaranteed.”
For more on-demand content from Mobile Edge Forum, go here.
