It’s no secret that containerization is spreading like wildfire throughout enterprise IT infrastructure. According to CNCF's Cloud Native Survey 2020, 92% of respondents are using containers in production. We know that containerization is here to stay, but there are still big questions outstanding on the best strategies for both adoption and orchestration. Given the leading role that containers will play in tech stacks over the long term, enterprises are debating the various options best suited for their needs.
Enterprises choosing to use container orchestration (Kubernetes especially) will always face the choice of whether to build their own solution or buy an existing package. There are some outstanding platforms offering complete solutions at the moment, but enterprises will sometimes be tempted by the DIY option in the belief that it will save money. Specifically, the concern is often around upfront cost — paying a platform service like Mirantis or Rancher feels like a big-ticket item, while the price of bare metal or public cloud look extremely low by comparison.
But while the idea of hiring a specialist or leaning on internal resources for container orchestration seems like a straightforward idea, the reality is that there has never been a worse time to do it yourself.
- Lack of specialized talent. If you’re even entertaining the decision between buying a ready-made solution and building your own, you probably don’t have the experts in-house needed to build a solution that will meet your needs. Containerization is not a challenge that can be addressed by generalists, and if you try to lean on them for your container orchestration, you’ll end up wasting a lot of time, money and talent. Even worse, your team’s attention will be pulled away from other projects, leading to delays in the features and products that are vital to your enterprise’s success.
- Deceptively low costs. When you’re using bare metal that you bought yourself, you can give your finance team an exact dollar cost of the upcoming deployment. However, the fact that four of your engineers are going to work on it full time, then burn out and quit – and you won’t be able to hire replacements – never appears on this balance sheet. This makes DIY Kubernetes Clusters appear incredibly cheap compared to platforms-as-a-service, but hopefully the error is clear from this description.
- Recruiting headaches. Maybe you’ve recognized that you don’t have the experts on staff to build your own container solution, but the right specialist could make the difference and still save money in the long term. If you try to find the right hire now, you’ll quickly discover what tech recruiters have known for over a year: Kubernetes talent is impossible to find. Not only is there a shortage of qualified candidates, but those companies that already have in-house Kubernetes experts are paying top dollar to keep them happy.
- Indispensable employees. So let’s say you are one of the lucky ones to have outstanding Kubernetes specialists on your team. Given their importance to the organization and the difficulty you’d face in trying to replace them, building your own platform means that your success is in the hands of two or three people. Those two or three people may one day pursue other opportunities, taking the institution’s entire Kubernetes knowledge base with them. When you’re stuck between a rock and a hard place of either paying increasingly large salaries to your Kubernetes experts or competing on the open market for a shallow pool of talent, are you really saving in the long-term over a Kubernetes-as-a-Service platform? What’s more, the subscription solution means that your tech stack will survive even if your retention leaves something to be desired.
Even after you’ve established the clear advantages of paying for an existing container solution, some finance teams will still balk at the fact that it’s impossible to guarantee costs in advance, given the consumption-based pricing model of many SaaS providers. While it’s true your costs will rise in line with consumption, here’s the other side of the equation you need to explain to them. With a pay-for-use architecture such as AWS Lambda, you know that you’ll have all the capacity you need during spikes, removing the risk of far more costly potential headaches.
For the vast majority of enterprises, the bottom line is that public cloud Kubernetes-as-a-Service offerings will save money and headaches in the long term. SaaS tools are well placed to offer the stability, security and observability assurances that DIY orchestration solutions cannot.