Run the database.
Don't run the database server.
Fully managed Postgres, MySQL, MongoDB, Redis, Kafka, and OpenSearch. Point-in-time recovery, automated minor upgrades, regional data residency. Connect with your existing driver — no proprietary lock-in, no transition fees if you leave.
Size the database. HA topology, plain pricing.
Every replica is billed transparently — no "Multi-AZ multiplier" that doubles your bill while the dashboard shows one resource. Backups bill on actual storage, not on a reserved ceiling.
Each node holds a full copy of your data. Storage scales up to 16 TB per node, sharded beyond that on Mongo / Kafka.
Backups are stored in object storage at $0.04/GB·mo. Point-in-time recovery (below) extends this to any second within the window.
Each replica is billed at the same hourly rate as the primary. Cross-region replication adds $0.005/GB of WAL traffic.
Six engines, none of them forked.
Every database runs upstream open-source code on the version you pick. We push patches back, we don't rename APIs, and your connection string works with the official driver.
OLTP, geospatial, vector search, time-series, anything SQL
Drop-in for legacy apps, WordPress / Magento / Shopify backends
Schema-flexible apps, content stores, IoT ingestion
Caching, session stores, rate limits, queues, pub/sub
CDC, event sourcing, real-time analytics, log aggregation
Full-text search, log analytics, observability, vector retrieval
How many nodes do you really need?
- Uptime
- 99.5 %
- RPO
- 24 h (last daily backup)
- RTO
- 15–30 min
- Uptime
- 99.99 %
- RPO
- 0 s (sync)
- RTO
- < 60 s automatic
- Uptime
- 99.999 %
- RPO
- 0 s
- RTO
- < 15 s automatic
What you don't have to think about anymore.
Daily full backup plus continuous WAL streaming. Retention from 1 to 35 days. Restore creates a new database — your live data is never touched.
Restore to any second within retention. Cross-region restore. Practice runs are free — we encourage them.
Minor versions auto-patched in your maintenance window. Major versions are a one-click switchover with replica-first cutover.
PgBouncer for Postgres, ProxySQL for MySQL, mongos for Mongo. Pooler scales with you; no separate billing line.
pg_stat_statements / sys schema / explain dashboards built in. Top 100 queries surfaced with auto-suggested indexes.
Prometheus endpoint, Grafana templates, OpenMetrics for everything else. 1-second resolution on CPU, IO, query rate.
Per-metric thresholds with webhook, PagerDuty, Opsgenie, Slack, or email. Throttling and incident grouping included.
Async replicas in any live region. Built-in load-balancing across replicas. Promote any replica to primary in under 30 s.
AES-256 at rest with customer-managed keys via NimbusNexus KMS. AWS / GCP KMS via federation for migrations.
Every connect, query, and DDL streamed to object storage or your own Kafka. Log schemas shaped for the evidence SOC 2, HIPAA and GDPR reviews ask for.
Private endpoint by default. Public endpoint behind allow-list if you need it. Same-region traffic to your VMs is free.
Backups, replication, and reads from your VMs in the same region are all $0. No "data transfer out" line item.
What teams typically stack together.
OLTP, multi-tenant schema, pgvector for AI features, pg_partman for soft-delete tables. Most teams here run Primary + replica in US-East.
Shopify-style storefronts, WooCommerce, Magento serving the North-American market. ProxySQL splits reads to replicas; group replication for write availability.
Container moves, customs events, warehouse scans streamed off Postgres via Debezium into Kafka topics — fanned out to track-and-trace, dispatch, and analytics. Singapore region keeps lag low to ports across SEA; tiered storage offloads cold partitions to object.
Session tokens, presence, and rate-limit counters for a West-Coast streaming app. Sentinel for failover, cluster mode for sharding beyond 64 GB. LA region keeps p99 RTT under 10 ms for creators on US-West.
Flexible schema for headless CMS, customer telemetry, GDPR-resident user records. Replica set for HA, sharding when one node’s storage isn’t enough. All data stays in the EU; backups never leave Frankfurt.
Vector logs from VMs and pods, hot tier on Performance NVMe, warm tier on Standard, cold tier on object storage. k-NN plugin for semantic search across audit trails.
Production Postgres, primary + replica. Three invoices.
4 vCPU, 16 GB, 200 GB SSD, 14-day backup retention, one read replica. Virginia (US-East). Source: published price lists, May 2026.
| Line item | NimbusNexus db-4-16 | AWS RDS db.m6i | GCP Cloud SQL | Azure DB for PG |
|---|---|---|---|---|
| Primary 4 vCPU + 16 GB | $59.86 | $198.20 | $186.40 | $204.10 |
| Sync replica (Multi-AZ) | $59.86 | $198.20 | $186.40 | $204.10 |
| Async read replica | $59.86 | $198.20 | $186.40 | $204.10 |
| 200 GB × 3 nodes (SSD) | $96.00 | $84.00 | $78.00 | $90.00 |
| 14-day backups | $3.74 | $36.40 | $32.00 | $34.00 |
| Point-in-time recovery | $0.00 | $0.00 | $0.00 | $0.00 |
| Cross-AZ replication egress | $0.00 | $18.20 | $16.40 | $18.20 |
| Total / month | $279.32 | $733.20 | $685.60 | $754.50 |
The hyperscalers double the compute line for "Multi-AZ" (it’s just a sync replica) and meter every byte of replication traffic. We don’t.
The same DSN your app already speaks. No SDK swap.
# Standard connection stringpostgres://app:••••@pg-prod.us-east-1.nimbusnexus.net:5432/mydb?sslmode=require# With connection pooler (auto)postgres://app:••••@pg-prod.us-east-1.nimbusnexus.net:6432/mydb?sslmode=require
# Atlas-compatible URImongodb+srv://app:••••@mongo-prod.us-east-1.nimbusnexus.net/mydb?replicaSet=rs0&tls=true# JS driver — no proprietary clientconst client = new MongoClient(uri);await client.connect();
# Connect with rediss://rediss://default:••••@redis-prod.us-east-1.nimbusnexus.net:6380# Or from your app:const r = createClient({url: 'rediss://...'});
Managed database questions.
See it built end-to-end.
Hypothetical workloads, concrete sizes, the failure modes worth knowing about, and the cost math at every layer. No fictional companies — just the architecture we would recommend if you walked in and asked.
An honest decision guide for the four databases most teams actually run. Where each one shines, where it bites, and the boring failure modes nobody warns you about until 2 a.m.
Read the playbook →PLAYBOOK · SINGAPOREContainer moves, customs events, dispatch — the data plane for a freight platform serving APAC trade lanes. Singapore as the primary, Frankfurt for EU receivers, with the cost and latency math at every hop.
Read the playbook →PLAYBOOK · FRANKFURTValidator and full-node fleets, the snapshot rhythm that survives a chain reorg, and EU data residency under our GDPR DPA, which keeps a treasury desk inside the EU regulatory perimeter.
Read the playbook →PLAYBOOK · VIRGINIAL4s for sub-7B models, H100s for the rest. Per-second GPU billing, weights pinned in same-region object storage, and the Postgres + pgvector control plane that turns a model deployment into a deploy command.
Read the playbook →Move from RDS, Atlas, or Cloud SQL. Cutover in minutes.
Stream live changes from your existing database via logical replication (Postgres / MySQL), change streams (Mongo), or replication slot (Kafka). Cut over when the lag hits zero. We pay the source-side egress.