What is the changing paradigm for communications service providers in terms of software integration and deployment, particularly the concept of CI/CD? It comes down to three aspects.
More network software, deployed from core to edge and RAN. Kalyan Sundhar, vice president and general manager for Keysight Technologies’ wireless network access business, says that across the board, lines are blurring in terms of software concepts and their role in the network, particularly in the context of Open RAN. While greenfield market entrants like Dish Network are pushing the concept of CI/CD more aggressively across their networks, all telecom operators have been pursuing implementations of more software-centric networks for years now—they virtualized their network functions, implemented cloud RAN,and have partnered with various cloud providers to handle network workloads and bolster edge-cloud availability, often in conjunction with 5G services.
Speed is seen as key. Telcos want to move faster than they have been able to in the past, and putting software-based networks into place has set the stage for new agility. The cadence of feature releases from major vendors were typically ingested once or twice a year and required arduous and extensive testing because of the scope of the releases. “Customers … don’t want to wait for this big waterfall model where you’re going to get an [update] and then wait for six more months before you get the next feature—and sometimes it’s longer,” says Sundhar. “The pipeline for getting something is really shortened by doing CI/CD, because you’re continuously integrating newly developed features into the mainstream.” Now, Sundhar says, he has heard of CSPs’ software release cadences stepping up to anything from once a month to once to a week—which is a huge leap in speed. A blog post from DevOps platform company Copado touts Telecom Argentina achieving “CI/CD maturity” and moving from being able to handle one major software release per year to “constant deployment” and “often releasing more than a dozen deployments each day without any negative impact on the final customer.”
The ability to more rapidly benefit from innovation. When software updates aren’t a huge monolith, the ability to integrate small changes on a weekly or biweekly basis can result in rapid realization of benefits. The flip side of that, Sundhar says, is that you also have the option to more easily back out of changes. “That’s the other advantage. You also have the option of quickly reverting back if something doesn’t [work out],” he adds. “You could trial something [and say] okay, this is not quite ready. You can back off and go to the previous build.”
It’s important that such changes be small, incremental ones that don’t require top-to-bottom, end-to-end validation of the entire network system, though. “You cannot do what you did in 12 months in a one-month cycle,” Sundhar explains. “The reason it took 12 months, or six months, or nine months, was because … changes are coming from everything, and you’re trying to make sure that it’s validated end-to-end.” In a CI/CD model, if you isolate the change—to a specific interface, perhaps—then testing can become a bit less cumbersome, he says: Spot checks rather than exhaustive testing. Where it gets tricky, he continues, if when an operator wants to promoting multiple things at the same time: Upgrading an infrastructure version from multiple vendors at the same time, for instance, multiplies the possible risk of something going wrong. “The sweet is spot is … that you want to contain [changes] as much as possible, then it’s easier to do the upgrade,” he says.
“Customers want to go to the disaggregated world, because they want to absorb innovation faster. They want to be able to monetize innovation faster,” says Tibor Fabry-Asztalos, SVP of product development engineering for Dell Technologies’ Telecom Systems Business. That could mean the ability to adopt new content, features or security updates and push them into production more rapidly—an attractive proposition when it comes to agility. But that’s not to say there’s no tension among speed, potential innovation and risk. “If you talk to CTO side … they want it really frequently,” observed Fabry-Asztalos. “If you talk to the operations, obviously they’re more cautious, because they have to deal with the risk coming in. … There’s always a balance between how frequently you want to innovate or take advantage of capabilities coming, versus how much risk you are taking on the operations side.”
Looking for more insights on CI/CD and continuous testing? Register for our upcoming webinar on the topic and stay tuned for the accompanying editorial special report.
