# Cluster Architecture (Cloud-Native Postgres)
Traditional monolithic PostgreSQL databases tightly couple the **Compute**
engine (parsing SQL, executing queries) and the **Storage** engine (writing
pages to disk, managing the Write-Ahead Log).
The defining innovation of modern, horizontally scalable Postgres architectures
is the **decoupling of Compute and Storage**.
![[arch_cluster_overview.png|256]]
By separating these concerns over a high-speed network, cloud providers route
write operations directly to a distributed, fault-tolerant storage layer. This
eliminates traditional bottlenecks like full-page writes and checkpoints, while
allowing the compute nodes to remain stateless and scale "serverless-ly."
Various implementations exist, each optimizing for different workloads:
- [[Aurora]]: Amazon's log-structured storage, eliminating traditional
checkpoints.
- [[AlloyDB]]: Google's intelligent, ML-driven caching for hybrid
transactional/analytical processing (HTAP).
- [[Neon]]: A true serverless architecture offering scale-to-zero compute and
Git-like database branching.