You know how much I care about OpenStack and how deep I feel involved in its community. A recent experience in the Netherlands at a premier customer made me think about OpenStack as a whole.
As I said at a conference last year, probably the most involved consultant in Europe, and yet I’ve seen more failures than success. Most failures were due to the lack of expertise in OpenStack and deep knowledge of Linux, protocols and OpenSource in general. I initially thought that the skills I have are quite common and that there’s plenty of people capable of running an infrastructure, but apparently, I was wrong.
Even big brands have usually 1-3 great engineers and the others are on the average. This is not really a bad thing, but you have to have great skills to manage OpenStack and most of the time management can’t rely on a bunch of guys for their business. Many decided either to go to public clouds and some went back to VMWare because skills are easier to find. To be honest, as an entrepreneur, I can’t blame them.
Public clouds (AWS, Azure, Google, …) are easy to embrace and you don’t have to maintain hardware, storage, network and is very attractive to those customers where IT is not their core business. Public clouds might seem costly at the beginning, but if you look at the real TCO (including labour cost), then you find out is not that much.
And if you are concerned about your privacy, a good VMWare cluster is enough for most of the businesses.
Kubernetes quickly ramped up into developers radar in the last year. It’s “cool” and containers are great ways for developers to distribute their applications. At the end of the day, companies need to run their applications to make money or support their business. How they do it, they don’t really care.
In my humble opinion, Kubernetes is not mature yet, especially in the networking and storage, and still lacking multitenancy. But is slowly getting there. Kubernetes is not that simple to manage, but it’s way less complex than OpenStack … and you don’t depend on MySQL or RabbitMQ to operate (which is a real pain). So what’s the need for OpenStack then?
This is the question I’m asking myself. Probably the number of use cases for OpenStack is quite small now, mostly related to telco operators and NFV.
The only thing that Kubernetes is not capable of is Microsoft Windows app, but Microsoft has shown interest in porting its apps to Linux (see SQL Server for example), not mentioning they are actively contributing to Helm.
While I still love OpenStack, we need to face the evidence that the interest in OpenStack is slowing fading away. However, its legacy has been invaluable to me and the community as well. The “Software-Defined” revolution that OpenStack brought, as well as the mindset around automation is the base for the future steps of IT.