Softbank sees business case for widespread MEC deployment
When Softbank started looking at deploying multi-access edge computing (MEC) in Japan around ten years ago, it had in mind low-latency applications such as autonomous driving, or remote surgery. A decade on and those applications remain in the future. But in the meantime Softbank has discovered a sound business case for MEC, as Ashiq Khan, VP, Head of Cloud Infrastructure R&D Division, Softbank, explained at DTW Asia 2023.
“What we found out is that you should be deploying MEC for everything. And that's our model. We are looking now at hundreds of MEC sites. Because if you can serve a customer … locally, what is the point of sending the traffic all the way to very far central data centers?” asked Khan. Whether a telco is handling video content or gaming software, having computing close to the end user, “saves enormous amount of traffic flowing through your core transport network,” said Khan.
Indeed, MEC becomes particularly important for offloading traffic from the core network as communications service providers (CSPs) ramp up 5G. “You save a huge amount of investment ... on the transport network when you really deploy 5G,” said Khan.
At present, he explains, the peak speed throughput of 5G is 10 gigabits, according to Khan. If one cell tower has three cells, then each of them will transmit 10 gigabits, which amounts to 30 gigabits per base station. “Now multiply it with the number of base stations you have in your country and if it's 50,000, you can very easily figure out how much traffic there is with the central data center -- [even] a hyperscaler … will not be able to scale it.”
Laying the ground for MEC
But building a MEC infrastructure requires careful planning. Khan recommended several measures that telcos should take when building out MEC, including automating the infrastructure and the cloud layer from day one.
“At each layer, we put in automation from the ground up … because doing automation later on is very difficult.”
Another area of focus for Softbank was the reduction of variability in its network so that it could simplify automation, which included reducing the number of suppliers.
“We have two hardware suppliers. And I restricted [us to] only two cloud suppliers. That's [still] variability but at the same time, you're hedging your risk,” according to Khan.
Khan also emphasized the importance of avoiding on-site maintenance by making full use of micro service architecture. “Software should be able to [self]heal … and should not be affected by a single hardware failure. So that's the operational model we have. We have automated it and it's done from a central location.”
And of course, customer management and experience are key areas of development, particularly now that the company has largely automated its infrastructure and cloud layer. Currently Softbank does not have a huge number of MEC customers, admitted Khan. But the company is taking the opportunity to learn how to onboard customers automatically and is also providing dashboards to each enterprise customer so they can see, for example, how much CPU they're consuming, or much bandwidth they are getting.