Frankly, I still see this a "niche" market - albeiit a large one
1) DevOps, like is explained here, looks to work only for in-house developed applications. If you're an ISV, it will inevitably end at the test stages, because you won't like to be responsible for roll-outs on sytems you have no control over.
2) There's always the costs/benefits ratio to take into account. This model requires a lot of resources (and/or a level of expertise), which may be not available in many SMBs situations. Not everybody is Google or Amazon.
3) In my experience, it's a long time developers need to do on-call and firefighting. Many sysadmins can't see the time they can say "it's an application problem" and pass the hot potato to developers (which often find that it was a deployment/configuration issue... <G>)
That said, developer can build-in realiability, scalability and security up to a point. Even if infrastructure can be coded, it doesn't mean it will be a developer to do it. Real syadmin has been coding infrastructure for years. Youn don't need fancy new buzzwords and standards to do that, most devices and software have been automatable for years, and skilled people did it.
Sure, new tools make this even more powerful, but you still need someone who knows the nuisances of orchestrating a datacenter to perform it, only a few developers may have that knowledge. And good sysadmins always worked with other tech roles to understand what is the best way to integrate everything into a smooth, optimized running system.
It looks we are talikng about bad sysadmins here. Those who don't give a damn about what runs on their systems, as long as it doesn't troubles them, and they don't care if they didn't spend half an hour to optimize it to run faster and better. The ones who prefer to spend their time on simple repetitive tasks, so they don't have to learn anything new, take any "risk" automating something, and maybe be asked to do something more in the time they then have. The ones if there are performance issues, just buy more and/or expensive hardware to solve them. Those who setup their systems as it was still NT4 or Unix SRV4 era, because they never kept their skills up to date - and never update a configuration since the device has been installed. Those who will never allow any change to their preciousssss datacenter for that very reason.
Do we need those? No, if they become extincted like dinosaurs life will be better.
Good ones will be finally able to spend more time on the real valuable (and enjoyable) parts of their job, instead of having to put again an ISO into a server and apply all the patches.
But are operations going away? Of course no. Systems will still need to be monitored. Hardware will still die and need replacements. User needs will evolve, and there will be still configuration changes to review, approve, implement. And user will make mistakes, while security incidents will happen.
The landscape evolves, and someone needs to keep updated knowledge about what is going on, and future needs. GitHub learned the hard way hardware firmwares may need to be updated as well.
DevOps sounds like a new buzzword for practices good teams always implemeted, maybe with less standardization and less automation - a buzzword to sell new tools, training and outsourcing to the very people who will never understand how to master it properly, and risk just to spend a lot of money for nothing. The others. are already doing - within the available resources.
Biting the hand that feeds IT
