This post originally appeared on Wired.com. For the last couple of decades, administrators have been split into a few different buckets. There are lots of specializations, but most admins are generally network, system, storage, or database admins. I’ve been seeing some signals from the market that these specializations are going away for the best admins.
When I started Puppet Labs almost 8 years ago, one of my major goals was to provide tools that enabled people to do operational work without being forced to know every little detail of the technology they were using. The idea was that if you didn’t have to care how a packaging system worked or how to create filesystems, you could focus more on the business value of your work and spend a less of your brain power on implementation details.
From Assembly to C
I have always thought of this change as akin to the shift from assembly programming (where programmers have to care about things like how many registers a CPU has or where the read head on a hard drive is) to high-level programming (where you can focus on the functioning of the program and largely ignore what kind of CPU or hard drive you have). This goal is exactly why Puppet has a Resource Abstraction Layer - we don’t think you should have to know how to install packages, create users, mount filesystems, or change a user’s password on Red Hat, Debian, Windows, OS X, and FreeBSD. Sure, you should be able to know it, but you shouldn’t have to know it to get your job done. The shift to higher-level programming languages goes further as an analogy, though, because it was only when that happened that people could begin writing portable programs that ran on multiple computers and at multiple organizations. Similarly, Puppet was built to enable portable infrastructure that could be version controlled, shared, and reused around the organization. Why should you become an expert in managing Apache or OpenStack unless it’s core to your business? Puppet isn’t the only product focused on enabling this kind of refocusing for the administrator. In fact, virtualization as a whole is all about isolating the administrator from needing to care about hardware details (among many other benefits). My experience is that it’s clearly working: Everywhere I go, there are administrators who have been freed from needing to maintain an encyclopedia of technical bits, and as a result are stepping into more strategic roles for their organization, and are able to cover much larger ground, while staying operational and directly delivering real-world value.
A Different Kind of Admin
These people end up looking like a different kind of admin. Some people call them DevOps admins, others have called them cloud admins; I’m less concerned about what to call them than making sure they’ve got the tools they need to do their jobs. The critical point for me is that their broadened focus has meant that they’re beginning to step outside of their traditional silos; I see admins asking themselves how they can integrate network, storage, database, and system into a single application lifecycle. I see teams being told it’s no longer ok to have to hand off control of a project to different teams just because they rely on different technology. I love this. I love seeing better tools enable people to build teams to focus on their customer instead of their technology.
Threat or Opportunity?
Obviously, if you’re an administrator today, this is both a threat and an opportunity. If your career is built on being that specialist who only cares about one silo, then you’re in danger of being left behind as part of this transition. However, if you’re adopting these new tools, if you’re trying to push the technology down so you can grapple with the needs of the customer, whoever they are, then you’re perfectly positioned to continue growing and facing new challenges. Like all changes of this nature, it’s not a sudden, boolean change. You’re not going to go from not seeing this to being victim of it. This transition began years ago, probably when virtualization first started getting big 5 years ago, and it will probably take at least 5 or 10 more years to really permeate the market. And of course there will always be specialists; someone needs to know the details. There are still assembly programmers, and people who write disk drivers and memory controllers. However, there are probably about the same number of assembly programmers today as there were 30 years ago, while the total number of programmers has grown exponentially. This is exactly what I think will happen in the administrator world: In 10 years, we’ll have roughly the same number of database, storage, network, and system admins, but 10 or 100 times as many people will be doing operational work without having to step into that level of detail. Based on what we’re seeing with customers and partners, it looks like 2013 will be a pivotal year in the progression toward that new world. Do you see this trend arising at your company?
- Read Luke's Whither the System Tools?
- Read all about the state of DevOps today in the 2015 report, and learn how DevOps can boost overall organizational performance, and how it depends strongly on IT leadership for success.