Six engines · one bill

A production cluster in 60 seconds.
A bill you can read.

Managed PostgreSQL, MySQL, Valkey, MongoDB, Kafka and ClickHouse. Backups with point-in-time recovery, pooling and private networking in the plan price, HA from a flat +70%, and we migrate you off RDS free, with zero downtime.

Free migration, zero downtime14-day money-back guarantee
raff db create · us-eastcluster ready
primary
postgres 16 · 4 GB
standby
sync replica · HA
INCLUDED
daily backup + PITR
02:00 · restore to any minute
postgres://…@db-prod.internal.raff:5432
$ raff db create postgres --size 4gb --ha
Provisioning primary · us-east · NVMe
Attaching HA standby · sync replication
Backups scheduled · PITR enabled
● Ready in 38s · private endpoint issued
The problem

A managed database has two failure modes: 3 a.m. and the 1st of the month.

Either you run it yourself and it pages you at night, or you pay someone and the managed bill arrives with ten line items you didn't order. Raff was built against both.

The 3 a.m. problem · running it yourself
Failover drills & standby promotionautomatic
Nightly backups & restore testingdaily + tested
Minor version & security patchingzero-downtime
Connection pool tuningbuilt-in
Disk-full & bloat monitoringwatched 24/7
The 3 a.m. pageours, not yours

Every row on the left is a page you won't get. You keep the schema; we keep the pager.

The invoice problem · the same 4 GB Postgres, two bills
db.m5.large instance$125.60
Multi-AZ standby ×2$125.60
gp3 storage 100 GB$11.50
Provisioned IOPS$38.00
Backup storage overage$14.20
Data transfer out$41.40
RDS Proxy (pooling)$32.85
total$389.15
raff · postgres 4 GB · HA · everything$67.98

One line. Backups, PITR, pooling, private networking and egress are in the price, not on the bill, and HA is a flat +70%, not a silent second instance.

Six engines

Your whole data layer, one roof.

Relational, cache, documents, streams and analytics: same console, same flat pricing, same private fabric. No second vendor for Kafka, no third for ClickHouse.

PostgreSQL
14 – 16
The default for new apps
  • PITR: restore to any minute
  • PgBouncer pooling built in
  • Read replicas on demand
from$7.99/mo
MySQL
8.0
WordPress to warehouse-scale
  • Drop-in for the MySQL ecosystem
  • Automatic failover, zero config
  • Daily backups included
from$7.99/mo
Valkey
Redis-compatible
Cache, queues & sessions
  • Point-in-time RDB snapshots
  • Eviction policies per use case
  • Scoped per-app credentials
from$7.99/mo
MongoDB
7.0 · 8.0
Rolling out
Documents without Atlas tiers
  • Replica sets standard
  • Continuous backup
  • No per-operation pricing
from$9.99/mo
Kafka
3.7+
Rolling out
Streams without a second vendor
  • Single-broker dev to HA clusters
  • Topics, ACLs & schema registry
  • Flat per-broker pricing
from$19.99/mo
ClickHouse
24.x
Rolling out
Analytics at wire speed
  • Columnar OLAP, billions of rows
  • Tiered object storage
  • No per-query billing
from$15.99/mo
In the price

Everything they call an “add-on” is our default.

Eight things that show up as extra line items, paid tiers or surprise multipliers elsewhere. Here they're just how a database ships.

IncludedDaily backups + PITRthem: paid add-on
IncludedConnection pooling (PgBouncer)them: RDS Proxy $$
IncludedPrivate networkingthem: VPC config
IncludedEgress over the fabricthem: $0.09/GB
IncludedMetrics & query insightsthem: paid tier
IncludedMinor version upgradesthem: extended support fee
IncludedTLS + IP allowliststhem: extra config
IncludedSpend cap on by defaultthem: no hard cap
Standalone, but one click from your app

Works with anything. Snaps into Raff Apps.

Use it from any host over TLS with IP allowlists. And if your app runs on Raff Apps, one click creates a virtual link: a private, encrypted wire straight from the service to the cluster. No connection strings pasted around, no public hop, $0 egress.

Credentials injected, never copiedTraffic never leaves the fabricRevoke in one click
my-app · connect databaselinked
{ }webraff apps
apiraff apps
virtual link · TLS
postgresmanaged · HA
+ Link to app
DATABASE_URL injected into web, api · egress $0.00
Price it yourself

Same database. Three bills.

Pick a plan and watch the same cluster priced on Raff, AWS RDS and DigitalOcean, HA multipliers and typical extras included.

Engine
Plan
Raff$67.98/mo
HA, backups, pooling, egress in
AWS RDS$133/mo
Multi-AZ 2× + extras
DigitalOcean$130/mo
standby 2× + storage
You keep vs RDS~$65/mo

Competitor figures are modeled on each provider's published pricing (compute, storage, HA multipliers, typical egress and IOPS extras), gathered Jul 2026. Re-verify before relying on them.

Head to head

Raff DB vs AWS RDS, DigitalOcean and Supabase.

Where each provider lands on the parts of managed that usually cost extra. Supabase is Postgres-only by design.

Raff DBBestAWS RDSDigitalOceanSupabase
Engines under one roofAll 6Relational only5 enginesPostgres only
HA standby + failoverFrom +70%2× compute2× computeHigher tiers
PITR (restore to any minute)IncludedStorage billedIncludedPaid add-on
Connection poolingBuilt-inRDS Proxy $$PgBouncer incl.Included
Egress to your Raff app$0$0.09/GBMetered250 GB then paid
Private link to your app1 clickVPC setupVPC setupN/A
Query insightsIncludedPaid tierBasicIncluded
Billing dimensions1–26–1024+
Spend cap on by defaultYesNoAlerts onlyPartial
Zero-downtime migration helpFreeDMS (DIY)Self-serveSelf-serve
Runs on own cloudYesYesYesResells AWS
Starts at$7.99/mo~$25/mo$15/mo$25/mo

Competitor rows reflect each provider's published pricing, docs and default plan limits, gathered Jul 2026. Re-verify before relying on them.

Migration

Leave RDS without a maintenance window.

Continuous logical replication keeps Raff in sync with your source while it stays live. When lag reads zero, you cut over: in seconds, not a weekend.

1

Connect the source

Point Raff at your RDS, DigitalOcean, Supabase or self-hosted instance with a read-only replication user. Nothing changes on your side.

source: rds-prod · connected ✓
2

Sync continuously

Logical replication streams every write to your new Raff cluster while the source stays live. Watch the lag counter fall to zero.

replica lag: 0.0s · rows verified
3

Cut over in seconds

Flip the connection string, or the virtual link if your app is on Raff. Writes land on Raff; the old bill stops.

cutover: 14s · downtime: none
White-glove migration, free.
Moving from RDS, DigitalOcean, Supabase or a self-hosted box? An engineer runs the cutover with you.
Book a migration
Dev experience

See the slow query before your users do.

Query insights, live metrics and log tails are in the console and the CLI, not a paid observability tier. One command provisions, one shows you exactly which query is eating the CPU.

  • Query insights: slowest statements, ranked, free
  • Live metrics & log tail in console and CLI
  • Scoped credentials per app, rotated in a click
  • Terraform provider & REST API for everything
db-prod · query insightslast 1h
SELECT … FROM orders JOIN users …412ms
UPDATE inventory SET qty …118ms
SELECT … FROM sessions WHERE …46ms
INSERT INTO events …12ms
$ raff db metrics db-prod --live
cpu 12% · mem 41% · conns 34/200 · cache hit 99.2%
replica lag 0.0s · last backup 02:00 ✓
Reliability

Your data. Our pager. In the open.

A public status page for the database fleet, an SLA with credits, and restores we actually test, because a backup you've never restored is a rumor.

99.9%Uptime SLACredited if we miss it
<30sAutomatic failoverStandby promotion, no ticket
ISO 27001CertifiedEncrypted in transit and at rest
rafftechnologies.com/statusAll systems operational
PostgreSQL fleet99.99%
MySQL fleet99.99%
Valkey fleet99.98%
Backups & PITR100%
View the live status page
FAQ

The questions switchers ask.

Yes. pgvector, PostGIS and 60+ other extensions are available on every plan at no extra cost. Enable them from the console or with a single SQL command.

Yes. It speaks the standard MySQL wire protocol on port 3306, so existing connection strings, ORMs, WordPress and phpMyAdmin work unchanged. Point them at the new host and go.

Yes. Valkey forked from Redis 7.2 and keeps the RESP protocol and command set, so redis-py, ioredis, Jedis, go-redis and the rest connect with no code changes. Point them at the Valkey host and go.

Reach for ClickHouse when you run analytical queries, counts, sums and group-bys over very large, append-heavy tables like events, logs and metrics. For transactional app data with lots of small reads and writes, Postgres or MySQL is the better fit. Many teams run both.

Yes, open-source Apache Kafka 4.3 running in KRaft mode, speaking the standard Kafka protocol. Your existing producers, consumers and CLI tools work unchanged.

Postgres 14, 15 and 16. Minor version upgrades are applied automatically with zero downtime; major upgrades are a guided operation you trigger when ready.

Valkey is the Linux Foundation fork created when Redis changed its licence. It keeps the same protocol, commands and performance profile, but stays BSD-licensed and community-governed. For your app it behaves like Redis; for your legal team it stays open source.

HA plans run a three-node Galera cluster: every node holds the full dataset synchronously, and HAProxy routes writes to a healthy node. If a node fails, traffic moves automatically with no data loss, no ticket and no manual step.

No. Pricing is flat per cluster size, like every other Raff engine, and backups are included. A growing query volume never grows the bill.

We bill a flat plan price per cluster instead of metering ingress, egress and storage separately. You size the cluster, you see one line on the bill, and throughput within it does not generate surprise charges.

Postgres clusters archive WAL continuously, so you can restore to any minute in the retention window. Restores spin up as a new cluster, so you can inspect the data before promoting it.

Yes. MySQL 8 with InnoDB is the database WordPress is built for. Create the cluster, point wp-config.php at the private endpoint (or the public one with an IP allowlist), and daily backups cover the rest.

HA adds a synchronous standby with automatic failover for a flat percentage of the plan price, from +70% depending on the engine, shown upfront on the same single bill line. On RDS and DigitalOcean, enabling HA silently provisions a second instance billed at full rate.

HA adds a synchronous standby with automatic failover for a flat percentage of the plan price, from +70% depending on the engine, shown upfront on the same single bill line. On RDS and DigitalOcean, enabling HA silently provisions a second instance billed at full rate.

You can. RDB persistence means your data survives restarts and failovers, and nightly plus on-demand snapshots cover recovery, restored as a new instance or in place. For strictly ephemeral caching, use it purely as a cache with eviction policies.

HA adds a synchronous standby with automatic failover for a flat percentage of the plan price, from +70% depending on the engine, shown upfront on the same single bill line. On RDS and DigitalOcean, enabling HA silently provisions a second instance billed at full rate.

HA adds a synchronous standby with automatic failover for a flat percentage of the plan price, from +70% depending on the engine, shown upfront on the same single bill line. On RDS and DigitalOcean, enabling HA silently provisions a second instance billed at full rate.

HA adds a synchronous standby with automatic failover for a flat percentage of the plan price, from +70% depending on the engine, shown upfront on the same single bill line. On RDS and DigitalOcean, enabling HA silently provisions a second instance billed at full rate.

One line per cluster: engine plus size, with the HA percentage folded in if you enabled it. The engine’s protection features, private networking, metrics and fabric egress are inside the plan. The only possible second line is overage storage at a flat per-GB rate.

One line per cluster: engine plus size, with the HA percentage folded in if you enabled it. The engine’s protection features, private networking, metrics and fabric egress are inside the plan. The only possible second line is overage storage at a flat per-GB rate.

Yes, it is a standalone product. Connect from anywhere over TLS with IP allowlists and scoped credentials. If your app runs on Raff Apps, a one-click private link adds a private wire with $0 egress on top.

Kafka is private-network only by design: clients connect over the private fabric or a one-click private link from your Raff services, never the open internet. Exposing Kafka brokers publicly is an anti-pattern, so unlike our other engines there is no public endpoint. Producers and consumers authenticate with SCRAM credentials over TLS.

One line per cluster: engine plus size, with the HA percentage folded in if you enabled it. The engine’s protection features, private networking, metrics and fabric egress are inside the plan. The only possible second line is overage storage at a flat per-GB rate.

One line per cluster: engine plus size, with the HA percentage folded in if you enabled it. The engine’s protection features, private networking, metrics and fabric egress are inside the plan. The only possible second line is overage storage at a flat per-GB rate.

One line per cluster: engine plus size, with the HA percentage folded in if you enabled it. The engine’s protection features, private networking, metrics and fabric egress are inside the plan. The only possible second line is overage storage at a flat per-GB rate.

Yes, it is a standalone product. Connect from anywhere over TLS with IP allowlists and scoped credentials. If your app runs on Raff Apps, a one-click private link adds a private wire with $0 egress on top.

Yes, it is a standalone product. Connect from anywhere over TLS with IP allowlists and scoped credentials. If your app runs on Raff Apps, a one-click private link adds a private wire with $0 egress on top.

Yes, it is a standalone product. Connect from anywhere over TLS with IP allowlists and scoped credentials. If your app runs on Raff Apps, a one-click private link adds a private wire with $0 egress on top.

Your database, minus the drama.

Six engines, backups and pooling in the price, a bill you can read in one line. Provision a cluster in 60 seconds, or let us migrate yours for free.