Why 2023 Became a Turning Point for Microservices: Insights from Google and Amazon

2023 proved to be a challenging year for microservices, with the once‑celebrated approach facing serious criticism while cloud giants Amazon and Google spearhead a new wave of innovation.

Google’s Rethink of Microservices

In the paper “Towards Modern Development of Cloud Applications” led by Google engineer Michael Whittaker, the authors argue that microservices have long mixed logical boundaries (how code is written) with physical boundaries (how code is deployed), creating fundamental problems.

Google engineers propose building applications as a logical whole and delegating placement decisions to an automated runtime that adapts to workload demands and available resources. Their prototype reduced system latency by 15× and cut costs by up to 9×.

Amazon Prime Video: Dropping Microservices for a Monolith

Amazon’s Prime Video engineering team published a blog post claiming that, for video monitoring, a monolithic architecture outperformed microservices and serverless approaches, achieving up to 90% reduction in operational costs.

The original system used AWS Step Functions and Lambda to orchestrate thousands of video streams, but scaling limits capped throughput at 5% of expected load, making data transfer between components prohibitively expensive. By consolidating all components into a single process, the team improved scalability and cost efficiency, concluding that microservices and serverless can work at scale but must be chosen based on specific use cases.

Ruby‑on‑Rails creator David Heinemeier Hansson commented that Amazon’s shift illustrates how microservices can add unnecessary complexity, and serverless may exacerbate it.

Revealing Microservices: Hidden Drawbacks

The term “microservices” was coined by Peter Rodgers in 2005, evolving from SOA concepts. While the approach gained widespread adoption, Google’s paper lists several drawbacks:

Performance: Network serialization and remote calls can become bottlenecks.

Observability: Distributed interactions make error tracing difficult.

Management: Independent release schedules increase binary management overhead.

Fragile APIs: Once a microservice’s API is set, changes can break dependent services, leading to API bloat.

Disrupting Traditional Microservices: A New Model?

When The New Stack reported Amazon’s move, many noted that the architecture was not a classic monolith. Former AWS cloud strategy VP Adrian Cockcroft clarified that this represents a new kind of microservice architecture, where the build method is chosen based on anticipated scale.

Google’s research addresses this by delegating execution responsibilities to the runtime, delivering microservice benefits with higher performance and lower cost.

Infrastructure Needs Reconsideration in 2023

The year also saw a broader reassessment of foundational architecture concepts, with cloud computing under scrutiny. For example, 37signals purchased Dell servers and left the cloud, challenging the prevailing belief that external cloud services always simplify operations.

Founder David Heinemeier Hansson argued that cloud marketing often overpromises simplicity, and real‑world operations still require substantial personnel.

FinOps gained prominence as companies like DataDog faced $65 million cloud monitoring bills, prompting many to adopt cost‑control tools such as KubeCost.

These pressures led organizations to revisit microservice strategies alongside other architectural decisions.

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