Rethinking Microservices in 2023: From Google’s “Microservices 2.0” to Amazon’s Monolith Shift

2023 has become a turning point for microservices, with major players reconsidering the paradigm that has long been treated as the de‑facto standard for cloud‑native applications.

Google Says Our Microservices Are Wrong

Google engineers, led by Michael Whittaker, published a paper titled “Towards Modern Development of Cloud Applications” that criticises traditional microservice architectures for mixing logical and physical boundaries.

They propose a “Microservices 2.0” approach that treats the application as a logical whole while delegating deployment decisions to an automated runtime, reducing latency by 15× and cost by 9×.

Google advocate Kelsey Hightower emphasised that starting from well‑organised modular code allows the deployment architecture to be an implementation detail.

Encourage developers to write single‑process applications composed of logical components.

Defer physical distribution and execution challenges to runtime.

Deploy applications atomically.

Amazon Prime Video Team: Abandon Microservices, Adopt Monolith

In June, Amazon’s Prime Video team released a case study showing that replacing a serverless, microservice‑based monitoring system with a monolithic architecture cut operational costs by over 90 % and simplified scaling.

The original solution relied on AWS Step Functions and Lambda, but high concurrency caused account limits and excessive state‑transition charges.

By rebuilding the infrastructure as a monolith, the team achieved better performance, lower cost, and improved scalability for video‑stream quality monitoring.

Others Also Leaving Microservices Behind

Numerous mid‑size teams have followed suit. Uber reduced the number of microservices, consolidating them into larger services managed by 5‑10 engineers each, while Managed by Q merged many microservices back into a Django monolith after two years of high overhead.

These experiences highlight that microservices are not a universal silver bullet, especially for small engineering groups.

The Illusion of Microservice Prosperity: From Monolith to “Distributed Monolith”

Critiques point out performance penalties from network serialization, difficulty tracing bugs, management overhead of multiple binaries, and fragile APIs that cannot evolve without breaking dependents.

Often, teams split a monolith into loosely coupled packages without establishing clear domain models, resulting in tightly coupled “microservices” that fail to deliver the promised benefits.

Google’s New Take on Microservices

Google argues that microservices should match the scale of the problem; separating programming and deployment models lets developers focus on code while the runtime optimises cost‑effective execution.

Their research shows that delegating execution to the runtime can retain microservice benefits while delivering higher performance and lower cost.

A Year of Infrastructure Re‑thinking

Beyond microservices, cloud computing itself faced scrutiny. 37signals (Basecamp) bought Dell servers and left the cloud, while DHH criticised cloud marketing hype and highlighted the limited operational savings for many organisations.

FinOps gained prominence as companies sought to control soaring cloud bills, with some organisations facing multi‑million‑dollar monitoring costs.

These trends suggest that architects must reassess past decade‑long engineering decisions, including the widespread adoption of microservices.

Reference links:

https://thenewstack.io/year-in-review-was-2023-a-turning-point-for-microservices/

https://dl.acm.org/doi/10.1145/3593856.3595909?utm_source=thenewstack&utm_medium=website&utm_content=inline-mention&utm_campaign=platform

Promote civil discussion using 交流技术 , 职位内推 , and 行业探讨 as the main focus.