We are in the midst of a revolution in datacenter infrastructure. The cloud, SSDs and hyper-converged systems are joining with software-defined infrastructure and automation/orchestration approaches to challenge the admin on a skills and relevance basis. It’s fair to say that some classes of admin roles are about to be eliminated, while new opportunities will open up for the agile admin to thrive in the Brave New World of next generation datacenters.
To understand the implications for any particular admin role, we need to understand the impact of all these changes to that role’s skills requirements. The easiest way is to look at the traditional partitions of servers, networks and storage. Because of the impact of hyper-convergence and network automation, I’m going to reverse the “traditional” order of things and look at storage first.
Storage had been almost comatose from an innovation viewpoint for many years and the arrival of SSDs, with a tremendous boost in performance, caught IT by surprise. The last 5 years have seen more change in how storage is bought, configured and run than all of the previous three decades. The mainstay of many storage operations, the RAID array, is obsolescent and with it goes a whole lot of training and expertise. Instead, skills in (scale-out) NAS and object storage are in demand.
At the same time, storage is becoming more software-centric, as the software-defined (SDS) approach opens up. The implication of SDS is that almost all of storage control and management is executed in software running in virtual machines or containers. Most of this code is new, offering parallelism and scaling that we haven’t seen before. The data manger or storage admin of that new-age datacenter will be skilled in the new approaches and able and willing to respond to the rapidly evolving environment they will bring.
The challenges of creating an efficient hybrid cloud impact storage disproportionally. The issue is that data has to be in two places at once, since the transfer times and latencies between public and private clouds are far too high for a single data repository to handle. The data manager role, relatively new, will assume great importance as the hybrid cloud approach takes hold in the industry. Security, integrity and agility concerns mean that placing data in the right cloud is critical to efficient operation and to allowing cloud-bursting to take place.
Even though SDS will heavily automate data placement within either cloud, tools to optimize across clouds are just in the concept stage of development. Data managers are needed to ensure data is placed where compute will access it. As SDS evolves, this will increasingly automate, but the experience and skills to build policies for data management will grow out of the earlier manual efforts.
Moving on to networking, we are actually much further along the path to software-defined networking (SDN). Again, most services are moved into the virtual cloud in the server pool, with the switch looking more like bare silicon. The aim of SDN is to allow central IT admins to define policies that allow virtual networks to be built and configured by departmental staff.
The implied level of automation radically reduces efforts to configure clusters and cloud for the departments to play in. This reduces the network admin role in IT, especially when software vendors provide pre-tested template policies. At the same time, it simplifies the admin role at the departmental level, though not as much. An admin wanting to stay in the networking field will need to learn the new SDN skills and be agile and aggressive in a shrinking segment of IT staff.
Most enterprise IT admins are used to the automation from either hypervisor deployment or cloud prototyping. As with SDN, the “manual” labor is much reduced. Inevitably, this will migrate the job focus for the admin towards tuning orchestration and optimizing applications. Hyper-converged systems will grow substantially in the next few years, due to the downsizing of storage appliances caused by ultra-fast SSDs. Essentially, the optimum size for both servers and storage is identical.
This convergence is not just a hardware phenomenon, it will bring server and storage admin silos together, leaving data management as a key admin team, together with those admins involved in storage software and tuning.
We can’t leave out legacy systems in all of this. The current revolution in IT is making legacy holdouts untenable. None of the new approaches can be seriously bothered about assimilating the COBOL mainframe, so the writing is really on the wall for the whole legacy approach. Clearly, this should create a raft of admin and programmer requirements for the decade it will take to transition to new age approaches.
Admins face a very different future as we evolve “software-defined” and all that goes with it. Don’t panic, though. The changes will take place over the next five to ten years, allowing time to move on in skills, etc. The new age datacenter likely will have fewer admins, but will add in other areas to improve operational agility. Take the challenge!