Creating predictable services over unpredictable networks

Some software-defined networks are currently coming from the hyperscalers, said Muriel Médard, chief scientist at Steinwurf

Nancy Shemwell, chief operating officer at Trilogy Networks noted that some of the advanced services that will require low latency will be offered through a mix of networks, which can face a challenge from a technical perspective.

“If we look at the market today, there is an outpour of advanced services, and they’re going to start to require low latency. But if you look at the network, it’s really a mishmash of networks. Because you’re talking about fiber, you’re talking about copper, you’re talking about wireless. You have multiple generations, both public and private and satellite,” Shemwell said during a presentation at the Mobile Edge Forum 2022, adding that management of all of these elements is “extraordinarily” cumbersome and expensive.

“The underlying services are just not reliable enough,” provided Muriel Médard, NEC professor of Software Science and Engineering MIT and Chief Scientist at Steinwurf. “So what do you do? Do you just get multiple services and have them all delivered to you? Or do you actually try to make the service that you have better?”

Médard continued: “The right question to ask is: At a particular time, in a particular location, what are the resources that are available to a particular user for a particular application? … Is this somebody who’s just making a big massive download for something they’re going to use later, or is somebody who’s using something interactive.”

“If you look at another network that we use all the time, which is very reliable, which is part of our infrastructure, you look at the electricity, and so in the electricity domain, the network figures out basically what resources are available. And the electricity is, in effect, something which is completely fluid. And what we can do with coding, which is a way of mixing together data to basically render it fluid and much more manageable, is that then we can manage resources, the way that we’re used to managing resources in the in the grid, that that provides us power,” Médard said.

When asked by who had the motivation and incentives to solve this issue, Médard responded that the stakeholder who decide to solve the challenges will emerge as the winner of the equation. “But the question is who’s motivated to do it right now, the hyperscalers are really coming from the application level. And of course, their motivation is to make their applications run. Well, if you have a very delay sensitive application, and the delays are not being met, the customers are not going to use that application, the people who are selling the application are going to suffer from that. So you see, the hyperscalers are sort of coming from the application and going down in order to make their applications run well, you then have the network operators. What happens to a large extent is the network operators have been having to service the applications but not necessarily getting the full benefit of it, but bearing the brunt of a lot of the provisioning that they have to do,” she said. “And they have, to a large extent, relied on the equipment providers. Now, the equipment providers are of course, very motivated to provide high quality equipment, but they’re not necessarily that incentivized to reduce the overall use of bandwidth by their clients. You know, they don’t pay for that directly. That’s really in the client’s wheelhouse exactly how they manage that. So there’s been a little bit of paralysis there, to some extent the operators are the most incentivize, but then not necessarily the most enabled.”

“What we’ve been seeing sort of emerge in the nexus of those three large components of the market, which are all essential components, is SD-WAN. And if you look at SD-WAN, some of the software defined networking is coming from the hyperscalers. Some of it is actually coming from the equipment providers, and some of them are standalone or very heavily connected in deep partnerships, or with some of the operators, I think, SD-WAN is sort of fulfilling that that gap,” Medard said. A Software-defined Wide Area Network (SD-WAN) is a virtual WAN architecture that allows enterprises to leverage any combination of transport services — including MPLS, LTE and broadband internet services — to securely connect users to applications