Key Software Architecture Trends for 2020: Micro Frontends, AsyncAPI, Data Mesh, Policy as Code, Serverless, Low‑Code, and More

Good software architecture aims to help manage complex systems, and recent innovations for distributed systems, event‑driven architectures, and big data seek to apply emerging best practices while steering engineers away from common pitfalls.

InfoQ’s Architecture & Design topic map highlights the main architectural concepts and their industry adoption status; the editors track new trends, note which ideas have crossed the chasm, and annotate the map with observations from their internal discussions.

The discussion identifies four new trends championed by innovators: micro frontends, AsyncAPI, data mesh, and policy‑as‑code.

Micro Frontends

Micro frontends aim to bring the benefits of microservices to the UI layer, allowing large applications to be broken into smaller, independently deployable pieces.

Luca Mezzalira (author of Building Micro Frontends ) notes that micro frontends are not a brand‑new trend but gained traction in 2019, that best practices are still evolving, and that they are an excellent complement to single‑page and server‑side‑rendered applications, especially for large teams working on shared domains.

AsyncAPI

AsyncAPI addresses the mismatch between RESTful APIs and event‑driven architectures, enabling a shift from synchronous request/response APIs to APIs built specifically for asynchronous communication.

Daniel Bryant observes that many organizations adopting event‑driven architectures are actively looking for an AsyncAPI‑like specification to standardize their asynchronous interfaces.

Data Mesh

Data mesh, introduced by Zhamak Dehghani of ThoughtWorks, promotes domain‑oriented data ownership to avoid the pitfalls of monolithic data warehouses.

Dehghani explains that data mesh is a response to the centralized data management model, aiming to enable organizations to become truly data‑driven by decentralizing analytical data ownership. Thomas Betts adds that tracking data architecture is becoming as important as modularizing data in large‑scale systems.

Policy as Code

Policy as Code extends the Infrastructure‑as‑Code paradigm to the architectural level, allowing desired system states to be expressed and enforced programmatically.

Daniel Bryant highlights that policy‑as‑code discussions are prominent at events like KubeCon and among cloud providers.

Early Adopters – Serverless

Serverless remains a debated topic; many view it as not yet having crossed the chasm, often seen as a reaction against microservices.

Charles Humble and Daniel Bryant both express skepticism that serverless has achieved mainstream adoption, noting that only specific functions (e.g., AWS Lambda) may have crossed. Jan Stenberg points out that even a simple “Hello World” serverless app is highly distributed and requires skilled developers.

Low‑Code/No‑Code Platforms

Low‑code and no‑code solutions promise to empower end‑users to customize systems without developer involvement, though industry veterans remain cautious about vendor‑driven hype.

Humble remarks that platforms like Microsoft PowerApps and Google’s acquisition of AppSheet indicate growing commercial interest, while Betts stresses that integration challenges, not platform building, are the primary concern.

GraphQL

GraphQL has clearly crossed the chasm and is now influencing large‑scale architectural decisions, similar to the adoption trajectory of TypeScript.

Humble and Betts both affirm that GraphQL’s adoption has moved from a niche API layer to a core design consideration across systems.

Software Architecture Ethics

The editors debate whether ethics should be tracked on the architecture topic map, concluding that it belongs more to software culture discussions but remains essential for technical leaders.

Humble emphasizes the importance of continuing ethical conversations, while Betts argues that ethical considerations must be guided by practitioners who truly understand them.

Other Topics

Reactive programming, HTTP/2, and gRPC have also crossed the chasm, whereas many other ideas progress more slowly compared to programming‑language trends.