Interview Arm thinks its architecture is primed to win enterprise workloads thanks to a combination of AWS, cloud-native development and open-source software.
That's the view of Chris Bergey, the chip designer's senior veep and general manager for infrastructure line of business, who last week chatted with The Register about the company's plans to evolve from a promising upstart to a viable enterprise alternative.
Bergey acknowledges that Arm's basic proposition of lower power consumption has not sparked widespread adoption. But he thinks the times have now come to suit the company, because cloud-native development and the containers it favours are well suited to multi-core CPUs that can dedicate one core to each chore. Amazon Web Services' Arm-based Graviton processors, now powering EC2 instances, make it easier to experiment with Arm by removing the need for risky capital expenditure on servers made by – let's be polite – the emerging vendors that have to date made Arm-powered servers.
Bergey rates open-source developers as the key piece, because if they ensure their projects work on Arm it makes life easier for the Arm-curious. He thinks FOSS momentum is now surging, citing recent Arm-adoption efforts at projects like PHP and Memcached as good signs that businesses will soon be able to run the code they already use on Arm hardware.
Once users see that Arm is an alternative that won't require major change to their code, he thinks the lower cost of running Arm CPUs kicks in.
"If you look at the CIOs and CEOs and CFOs for whom cloud operational expenses is one of their dominant costs, for AWS to say its Arm instances offer 40 percent savings is pretty compelling."
Bespoke applications are another opportunity. "We have been very focused on the cloud native ecosystem," Bergey said. "We think it is approaching 50 per cent of workloads, especially in new deployments."
Bergey also thinks that the proliferation of devices helps Arm.
"As you get to the edge the edge physical environments are more challenging and power efficiency can mean smaller battery or convective cooling," he says. Cloud-native also makes sense on the edge, because it is easier and less disruptive to automate delivery of an updated container that improves a device's functionality or security than it is to update firmware.
"Things like an old Linux kernel in a router are solvable with cloud-native," he says, and that's important because users of industrial devices and telecoms infrastructure "don't want a truck roll every time they make a change."
Arm also thinks servers running its kit have potential, but Bergey held his cards close to his chest on that topic.
"I can't talk about everything in flight with servers," he said. "A lot of the servers players are servicing the market that has friction for making the move." Those players will, he believes, pivot to growth markets.
"As Moore's Law slows, you need to solve things at the system level," he said, offering the move from three-tiered computing to hyperconverged infrastructure as an example of a market changing rapidly.
Bergey is paid to be optimistic about his company's products. But also realistic about Arm's prospects.
"We are seeding things that are a few years from revenue," he admitted, but also hopes the company is spinning up a "flywheel of the larger ecosystem".
Which is what Intel has done for decades, with extensive assistance for software developers and long-range planning to bake useful features into silicon to make sure that emerging workloads get the power they need.
Arm's approach and reliance on open-source and clouds is different. But so are the times. ®