Platform Engineering

DevOps asked development teams to take on more of the work of running their software: infrastructure, pipelines, security, monitoring and on-call. In large organizations, that created a new problem. Every team had to learn and maintain a growing pile of cloud services and tools, and much of that effort was duplicated across dozens of teams. Platform engineering responds by building shared internal platforms that handle the common, undifferentiated work. The aim is to give every team the benefits of DevOps without each team needing deep infrastructure expertise. Developers use self-service tools and well-supported defaults to create services, deploy and observe them, while a dedicated platform team keeps the underlying infrastructure secure, reliable and up to date.

Platform engineering grew in prominence in the early 2020s, building on ideas from site reliability engineering, DevOps and the platform team concept in Matthew Skelton and Manuel Pais's 2019 book Team Topologies. Industry analysts have predicted that most large software organizations will have platform teams, and open source tools such as Backstage, released by Spotify in 2020, have made developer portals common. For engineering leaders, infrastructure and DevOps teams and developers, platform engineering is a way to make good practices the easy path rather than an extra burden.

Definition of Platform Engineering

Platform engineering is the practice of designing, building and operating an internal developer platform as a product, providing curated self-service capabilities, often called golden paths, that reduce the cognitive load on development teams and standardize how software is built, deployed, secured and operated.

Its main output is an internal developer platform, it builds on DevOps practices and it aims to improve developer experience (DevEx).

Core Ideas

Platform as a product. The platform team treats developers as customers, researching their needs, measuring adoption and improving the platform based on feedback.

Self-service. Developers can create environments, services and pipelines on demand, without filing tickets and waiting for another team.

Golden paths. The platform offers recommended, well-supported ways of doing common tasks, such as creating a new service, so the easy path is also the secure, compliant path.

Reducing cognitive load. The platform hides complexity developers do not need to manage, letting them focus on building products.

Platform Engineering and Related Disciplines

DevOps describes a culture of shared ownership between development and operations. Platform engineering provides the tools and services that make that culture practical at scale.

Site reliability engineering focuses on the reliability of production systems. Platform teams often work closely with SRE teams, and in some organizations they are combined.

Traditional infrastructure teams respond to requests. Platform teams build products that let developers serve themselves. In Team Topologies terms, a platform team is one of four team types, existing to reduce the cognitive load on stream-aligned teams that deliver products.

Importance of Platform Engineering in Delivery and Engineering Operations

Platform engineering speeds up delivery. Self-service infrastructure and standard pipelines remove the waiting and setup that slow new work.

It improves consistency and security. When most teams use the same golden paths, security controls, monitoring and compliance checks are applied everywhere by default.

It also improves developer experience. Developers spend less time on repetitive infrastructure tasks and more time on the work they were hired to do.

Platform Engineering and Financial Management

A shared platform avoids each team building and maintaining its own tooling, reducing duplicated effort across the organization.

Standard infrastructure templates can include sensible cost controls, such as automatic shutdown of test environments, supporting cloud cost optimization.

Platform teams are a real investment in people and tools, so their value should be measured through adoption, delivery speed and time saved for developers.

A well-run platform makes it easier to allocate cloud and tooling costs to the teams and products that use them.

Common Challenges

Platforms built without talking to developers often go unused, as teams find workarounds that better suit their needs.

Mandating a platform before it is good enough breeds resentment. Platforms should earn adoption by being the easiest option.

Platform teams can become bottlenecks if they try to support every possible use case instead of focusing on the most common needs. Measuring a platform team by the number of features it builds, rather than by developer outcomes, leads to platforms nobody wanted.