Compose an enduring DevOps organization structure
A comprehensive collection of articles, videos and more, hand-picked by our editors
NEW YORK -- Changing IT models call for different staffing needs, but that is easier said than done for large...
By submitting your personal information, you agree that TechTarget and its partners may contact you regarding relevant content, products and special offers.
One of the biggest challenges to shift to a DevOps job model is to find people with the skill sets and intellectual curiosity needed to have a hand in all layers of the stack. And with the growing adoption of public cloud and emerging trends, such as microservices and containers, it's as much about the culture as it is about the technology.
This issue of changing demands for IT hiring was discussed at the Open Network User Group conference here last week, with leaders from the financial sector discussing their own approaches to this shift.
"It's an evolution, if you like it or not," said Harmen Van der Linde, CitiCloud solutions manager and senior vice president for emerging networking and storage technologies at Citigroup. "It's a bit naive to assume you can just start from scratch and start all over and do it correctly from day one."
Citigroup takes a two-pronged approach where the broader initiative tries to combine designs around converged infrastructure, while green field projects are tested and slowly integrated into the larger system when appropriate.
Aditya Akella, an associate professor at the University of Wisconsin-Madison, looked at nearly 900 data center networks to get a better sense for how changes impact legacy workloads. Akella found certain things, such as adding a VLAN, can easily be automated in a new system like software defined networking, but actions that touch multiple layers of a hierarchical infrastructure network still require a much more intricate set up that can largely only be done manually.
There is a contract between the network and the application, Akella said. Some things the application expects the network to do and some things the network absolves itself from, and the fault tolerance and various boundaries vary.
"The level of automation or whatever programming you need is largely driven by where that contract is drawn between application and the network," Akella said.
There's a balance between addressing technical problems and organizational problems, as there is sometimes an inherent lack of trust between groups within a company, said Steve Russell, managing director at Morgan Stanley.
"One way to solve that is to get them together, but part of the change is a cultural change of what it is OK to change dynamically," Russell said. "The networking bar for that has historically been low and the server bar has been much higher."
DevOps jobs require a holistic view
A big challenge that comes along with this transition is to find engineers who understand the full stack and limit the number of specialized employees, Van der Linde said.
"What we need to get to is where folks have a full set of skills that can span network, storage and compute," Van der Linde said.
These cloud architects must have a holistic view across sufficient domains, but the last thing you want to do is educate the application about the infrastructure, Van der Linde said. You can end up with more tickets and app developers who think they know better than the people who provide the resources, so it's best to provide a high level of application abstraction from the resources.
"Instead of saying 'Tell me what size VM or storage you need,' what we are saying is 'Tell me what application stack you need,'" Van der Linde said. "From there we'll be able to stand up design pattern and do all the engineering qualification to figure out what it means from an IT resources perspective."
What most people struggle with culturally when they shift traditional IT roles into DevOps jobs is the characteristics of the workflow, Russell said.
"The people doing DevOps, they're sort of doing it horizontal to the layer they are used to operating in or the layer immediately below them that they are frustrated with," Russell said.
There is a shift to this full-stack integrator with the capabilities needed to improve time-to-market services, but that doesn't mean there won't be a need for deep knowledge in specific areas, said Pablo Espinosa, director of network engineering at Intuit.
"I don't necessarily see a massive shift there, rather than an expansion or extension of the role," Espinosa said.
Trevor Jones is a news writer with TechTarget's data center and virtualization media group. Contact him at firstname.lastname@example.org.
Understanding the relationship between DevOps and cloud
Boost cloud app security with a rugged DevOps approach
The specific skills every cloud architect needs