Software-defined networking used to be a term with specific meaning, but now "thanks to our friends in marketing," it's "a completely meaningless term," says Huawei's David Lenrow.
Speaking at Light Reading's NFV & Carrier SDN conference in Denver last week, the Huawei Technologies Co. Ltd. chief architect of SDN and and distinguished engineer proposed a new working definition for the oft-used term -- one that's based on an "aspirational goal."
"I think what software-defined means is to adopt the agility of computing outside of computing, and how we get there," Lenrow said.
However, getting there requires crossing hurdles, mainly the telcos themselves, Lenrow said.
"Historically telcos are risk-adverse, with huge billion-dollar-generating networks that just can't go down. Risk is the enemy," he said.
And everything requires a hardware forklift, including configuring features, which is a proprietary process.
Network operators are tempted by SDN, but see a need to spend millions of dollars to get the capabilities they already have. They need to "forklift their entire hardware." And "if it doesn't work [they're] in big trouble," Lenrow said.
Other barriers to telcos adopting the agility of the computing network include a data plane that is not programmable, integration is implementation-specific, and solutions are vertically integrated so that if a service provider wants to try a new feature that is not offered by its vendor, they have to rip and replace the whole stack, noted Lenrow. "If we can remove those three barriers, we'd have a lot more people crossing the chasm."
But all is not lost. There is good news. The industry is jumping hurdles at last.
Data plane programmability success is on the horizon. "Lots of startups, all the big vendors -- everyone's working on it," Lenrow said, adding that the P4 programming language is a promising new development.
Software and hardware disaggregation is also making progress. This is an important step in reducing cost and risk, and making change easier, noted Lenrow. "This is a conversation that needs to happen because it not only reduces cost but because it makes it safer to try new things."
Finally, intent-based networking is another way to ease the challenge of vendor-specific networks. "Traditionally, the way you configure a network is you type in a description of how you want the network to behave in terms of protocols," he said. "With the intent-based approach, we say, 'Instead of building a model of the network, let's build a model of the application -- and what it needs in terms of end-to-end connectivity, QoS, etc., and let's get a smart piece of software to figure out how to translate that.'"
As networks become more dynamic, service providers will need to shift to intent-based networking that describes the problem, not the solution, and lets SDN controllers make smart decisions. "Tell me what you need and not what to do," said Lenrow. "A big part of it is letting these smart controllers make smart decisions. If you tell them what to do in extreme detail, they are going to be inefficient and poorly optimized."
Intent-based networking is intended to "level the playing field and eliminate the vendor lock," he added. "Operators are the primary beneficiaries -- and the reason it's not suicide to do this as a vendor is because the network effect will drive the ecosystem and lead to a virtuous cycle."
— Elizabeth Miller Coyne, Managing Editor, Light Reading