Beacon blew up AWS bills: a 700% spike on RDS for PostgreSQL forced a pivot to self-hosted infra. The culprit wasn’t a misconfiguration; it was the price of managed persistence ballooning a side project's costs. [1]
Beacon’s answer wasn’t a cloud-agnostic reset, but a move toward a lean, self-hosted stack. A Raspberry Pi runs the core services, while a homegrown deployment/monitoring agent handles updates and alerts. The usual suspects — Grafana/Prometheus for metrics; OpenSearch/ELK for logs; and Metabase dashboards — feel heavy on tiny devices. [1] It’s cost control with a side of ops debt, but it’s cutting the bill in a tangible way.
On the ops side, Multigres shows a concrete path for scale: horizontally scalable multi-tenant PostgreSQL architectures aim to keep tenants fast and isolated without a monolithic upgrade. [2]
Meanwhile, the search debate tilts away from Elasticsearch. Build blazing-fast full-text search right in PostgreSQL with Supabase is viable for many apps; small-to-mid apps can skip Elasticsearch. If you’re chasing lighter engines, Meilisearch offers a friendlier alternative. [3]
Bottom line: teams are leaning toward self-hosted or tightly integrated Postgres setups to trade maintenance heft for cost control and performance.
References
Show HN: Beacon (open source) – Built after AWS billed me 700% more for RDS
AWS billing spike from RDS/Postgres charges; pivot to self-hosted infra; monitoring tool choices; open source beacon project.
View sourceMultigres: Horizontally scalable multi-tenant Postgres architecture
Post describes Multigres, a horizontally scalable, multi-tenant Postgres architecture for managing tenants while scaling reads and writes across instances efficiently.
View sourceSkip Elasticsearch: Build Fast Full-Text Search Right in Supabase
Advocates Postgres FTS with Supabase as an alternative to Elasticsearch; compares to Meilisearch and other engines, debates scale tradeoffs.
View source