“Platform Engineering Team” is not right for your company
How I would organize a large engineering team today, is by going back to the original idea of DevOps: “You build it, you run it”.
Until very recently, I myself would consider that idea unrealistic: the knowledge of the tools and skillset required to “run it” was both significant and different to what is required to “build it” so people had to specialize.
There emerged the concept of “Platform Engineering” where a special platform team builds self-service workflows that frontend and backend engineers use to build, deploy, monitor the applications.? That way individual product teams still seemingly own their apps, but the complexity of infra management is abstracted away.
Tools and Platforms
The main reason Platform Team is not a good idea is that tools available on the market have gotten a lot better very quickly.
AI is playing a role in that too, but just on its own what you get from the three major cloud providers and platforms like GitHub, Grafana is simply enough.
You do absolutely need to have someone who understands how it all works, but it doesn’t need to be a full time job. Senior backend or full stack engineers know the stack well enough to do what’s needed.
Even if it is a separate role, it is completely possible and preferable to have those “devops experts” embedded into the product business units, on some level, as opposed to being in a separate organization.
领英推荐
No, teams need to communicate with each other and share what they build. Just like with the open source model: what is best will get used most.
People can figure it out. Just like thousands of SOC2-compliant startups and small teams out there have. It is not rocket science. The more teams are made responsible for compliance of their stuff, the more compliant and secure it will actually be.
Whatever cost optimization you might aim to achieve with centralized tooling (or infra) it will be annihilated by the hidden costs due to the loss of productivity and autonomy of the teams building products for the end customer. This one is not so obvious, but is my firm belief, that engineering teams, on some level, should own their budget and the consequences.
A computer engineer who likes to solve challenges and explore new things.
1 个月Infra tools are so good now that most teams don’t need a separate Platform Engineering team. Devs owning their own deployments makes things faster and cuts out bottlenecks. Compliance and security aren’t rocket science, teams can handle it just like startups do. Centralized platform teams often slow things down more than they help. Let teams manage their own infra and budgets, and they’ll be more efficient and accountable.
Engineering Manager
1 个月And why do you even need a team to describe a docker container??
LLM-AI / DevOps / Cloud
1 个月Thanks for a well written and and thought-provoking post. A couple random thoughts here. Communications grow at O(n^2) - the platform org can ensure comms act as a 'hub and spoke' reducing communications overheads for the eng teams (who no longer have to directly coordinate with every other eng team on a change). In a small org with a small number of teams/engineers direct communications is faster and more efficient. In a large org then a platform hub-and-spoke is the faster and more efficient option. If the technologies are mature the technical-leader can define the off-the-shelf techs which will comprise the platform "Thou shalt use kubernetes and Istio etc etc" then the teams can get by using the standard without using a platform team's custom tooling. But if the needs require (re)-inventing the wheel (netflix in 2010 wanting to use microservices) then a platform team has to do it.
Chief of Platform | Army Software Factory
1 个月"You do absolutely need to have someone who understands how it all works, but it doesn’t need to be a full time job. Senior backend or full stack engineers know the stack well enough to do what’s needed. " Don't hire a platform team...just hire a platform team ....
Senior Director of Engineering at Upwork
1 个月In my opinion, "Platform engineering" as a term fundamentally lacks a shared understanding and meaningful common ground. Unless companies actively challenge the status quo and balance innovation with necessary controls it's slow and hard to fully leverage the new advancements.