Programmable Infrastructure is Developer Driven
The New Stack’s pancake breakfast and podcast circuit joined the Software Circus last week at the Amsterdam Roest. The New Stack Analysts host Alex Williams explored the conference theme of “programmable infrastructure” with a panel that included David Blank-Edelman, technical evangelist for Apcera, as well as three of the conference keynote speakers: Adrian Cockcroft of Battery Ventures; Ken Owens, CTO Cloud Services at Cisco Systems; and Kelsey Hightower, product manager and chief advocate at CoreOS.
For more episodes, check out the podcast section of The New Stack.
This podcast is also available on YouTube.
Alex gets things rolling by asking the panel what’s at the forefront of their thoughts regarding programmable infrastructure.
Adrian suggests “developer-driven infrastructure” is another way of looking at this phenomenon. This advances the conversation to the current conditions and considerations of developers, who by default have been handed ownership of the expanding breadth of the development process, along with its security challenges and its crucial role in the marketplace.
The demand for speed, the growing complexity of the development pipeline, and the increasing attention to security means a new reality for how platforms are perceived.
“Developers are now concerned with how quickly they do things, how efficiently they do things, and how securely they do things,” Adrian said. “And they need platforms that can express those out to them.”
“They need help. They need systems.” Those systems should be API-driven platforms, he says.
David mentions that Apcera’s policy-first solution within its Hybrid Cloud OS provides rule sets that act as security “guardrails” — developers can work as fast as they want because they’re working with a net.
From his perspective as a sysadmin for many years, Kelsey’s opinion is that the infrastructure has “always been programmable, but the interfaces are terrible.” The standard libraries are what have been improved in recent times, he says, citing AWS, Apcera, CoreOS and Docker.
“It gets more complicated every year for developers,” says Ken. “It’s been very difficult to programmatically define an environment to build your applications on top of, because you have to understand the network pieces, you have to understand the storage pieces, and you have to understand the virtualization environment to some extent.”
While affirming that he’s a big fan of APIs, Ken says that these days developers need “access to quick services.”
“You need a network service: you need a storage service. Why define a bunch of APIs that require the developer to figure out?” says Ken. “Let’s just write the code, put the code out there and make money on it. If the network gets broken, call me and we’ll fix it — that’s what we do.”
Ken’s stance on APIs sets up much of the conversation going forward, with David and Kelsey providing rebuttals. Adrian notes the distinction between two issues, one being the long-term stability of APIs, and the other being their rapidly increasing number. Ken reiterates that developers simply need a service layer to help them avoid getting bogged down in API management.
Having observed this back-and-forth regarding APIs, Alex suggests that this is only part of the enormous complexity of these architectures, and the discussion turns toward keeping everything under one tent.
Apcera, Cisco, CoreOS and Docker are sponsors of The New Stack.