Stop babysitting
the database at 3 a.m.
Managed MySQL 5.6 / 5.7 / 8.0 with automatic primary-standby failover; daily automatic backups plus binlog retention for recovery to any second; read/write splitting proxy, parameter groups, slow log and monitoring alerts out of the box — provisioned in minutes.
How the high-availability edition works
Applications connect to one fixed private address; when the primary fails, the proxy switches to the standby within 30 seconds and nothing changes in your config.
The high-availability edition pairs a primary with a hot standby using semi-synchronous replication; when health checks detect a primary fault, the standby is promoted automatically and the connection address stays the same. The read/write splitting proxy sits on that same address and routes reads to read replicas by SQL type while writes go to the primary.
- Automatic backups once a day at a time you choose, kept 7 days by default and up to 30; binlogs are retained continuously for point-in-time recovery to the second.
- Up to 5 read replicas in the same VPC as the primary, with replication lag you can monitor.
- Instances are reachable only inside the VPC and gated by a whitelist; public access must be enabled explicitly.
Four numbers you can verify
Every cell can be checked in console monitoring or a recovery drill.
Primary-standby HA, same address
Primary and hot standby run in different availability zones with semi-synchronous replication; health checks promote the standby automatically, the connection address never changes, and the switch usually completes within 30 seconds.
14:03:19 promote standby → primary ✓
14:03:27 proxy rerouted · writes resumed (16s)
Daily full backup + binlog to any point in time
Automatic backups are kept 7 days by default and up to 30, with binlogs retained continuously; a dropped table or a bad release can be rolled back to the second before it happened, restored to a new instance without touching production.
Read/write splitting, reads scale out
The proxy routes by SQL automatically: reports and queries go to replicas, the primary handles only writes; add or remove replicas on demand.
RSSD storage, online expansion
Data lives on triple-replicated RSSD cloud disks, expandable online from 20 GB to 32 TB; instance type changes complete inside the maintenance window with the same address.
A database you can see
CPU, memory, disk, IOPS, connections, QPS, TPS, slow queries, replication lag and more — 12 metrics with threshold alerts pushed to you.
Instance types and architectures
Pick the architecture by workload first, then the instance type by connections and data size; every type can be resized.
| Type | vCPU | Memory | Max connections | Storage (RSSD) | Best for |
|---|---|---|---|---|---|
| 1C2G | 1 | 2 GB | 1,000 | 20 – 500 GB | Development, testing, personal projects |
| 2C4G | 2 | 4 GB | 2,000 | 20 – 2,000 GB | Websites, back offices, small application databases |
| 4C8GRecommended | 4 | 8 GB | 4,000 | 20 – 4,000 GB | E-commerce order databases, SaaS databases |
| 8C16G | 8 | 16 GB | 8,000 | 20 – 8,000 GB | High-concurrency databases |
| 8C32G | 8 | 32 GB | 10,000 | 20 – 16,000 GB | Large datasets with lots of hot data |
| 16C64G | 16 | 64 GB | 16,000 | 20 – 32,000 GB | Core transaction databases |
| 32C128G | 32 | 128 GB | 32,000 | 20 – 32,000 GB | Core databases of large platforms |
Max connections are parameter-group defaults and can be tuned by memory; types and regions on sale are as shown in the console.
| Architecture | Nodes | Availability | Failover | Read scaling | Best for |
|---|---|---|---|---|---|
| Single node | 1 primary | No automatic failover; relies on backup restore | — | — | Development, testing, non-critical workloads |
| High availabilityRecommended | Primary + hot standby, across zones | 99.95% SLA | Automatic within 30 s, same address | Read replicas optional | Production databases |
| HA + read replicas | Primary + standby + up to 5 replicas | 99.95% SLA | Automatic within 30 s | Proxy splits reads and writes | Read-heavy, report- and query-intensive workloads |
| Version | Engine | Character set | Parameter group | Notes |
|---|---|---|---|---|
| MySQL 8.0Recommended | InnoDB | utf8mb4 default | Default / custom | First choice for new workloads: window functions, CTEs, atomic DDL |
| MySQL 5.7 | InnoDB | utf8mb4 / utf8 | Default / custom | Migrating existing workloads, JSON type |
| MySQL 5.6 | InnoDB | utf8mb4 / utf8 | Default / custom | Legacy compatibility only; upgrade recommended |
| Version upgrades | — | — | — | 5.6 → 5.7 → 8.0 upgrades run inside the maintenance window with an automatic backup first |
Features
Each one maps to an action or a switch in the console.
Automatic backups and binlog All architectures
Daily full backups at a time you set, kept 7 to 30 days; binlogs retained continuously.
Point-in-time recovery All architectures
Restore to any second inside the retention window as a new instance, without touching production.
Read/write splitting proxy High availability
One fixed address, automatic routing by SQL, adjustable replica weights.
Parameter groups All architectures
Change max_connections, innodb_buffer_pool_size and more online; a few need a restart.
Monitoring and slow log All architectures
12 metrics plus slow query analysis, with threshold alerts pushed to you.
Whitelist and VPC-only access All architectures
Reachable only inside the VPC by default; public access must be enabled explicitly and restricted by source.
Resizing and expansion All architectures
Storage expands online; type changes complete inside the maintenance window with the same address.
Data migration All architectures
Import with mysqldump or a physical backup, or set up replication for an online migration.
OpenAPI and CLI All architectures
Creation, backups, restores and read replicas can all be automated.
Use cases
Four typical application databases, each with an architecture, an instance type and three essentials.
Application database for cloud hosts
Put the databases behind websites, back offices and APIs on managed MySQL in the same VPC as the application instances — no more installing MySQL, configuring replication or doing inspections yourself.
Cross-border e-commerce order database
Orders, products, inventory and settlement on the HA edition; scale up and back up before a sale, scale down afterwards; run reports on a read replica.
SaaS multi-tenant database
Split databases or instances by tenant size with parameter groups managed centrally; point-in-time recovery fixes a single tenant's mistake.
Read-heavy query platforms
Reports, query APIs and operations analytics run on read replicas while the primary focuses on writes; the proxy routes automatically with no code changes.
Pricing and billing
Billed by instance type plus storage capacity; the HA edition is priced for two nodes; unit prices are shown in the console.
Single node 2C4G
Development, testing, non-critical workloads
- Nodes
- 1
- Storage
- RSSD from 20 GB
- Backups
- 7 days free retention
- Billing
- Monthly / hourly / yearly
High availability 4C8G
E-commerce order databases, SaaS databases
- Nodes
- Primary + hot standby
- Storage
- RSSD from 100 GB
- Backups
- 7 days free retention
- Billing
- Monthly / hourly / yearly
High availability 8C16G + replica
High-concurrency databases, reporting queries
- Nodes
- Primary + standby + replica × 1
- Storage
- RSSD from 500 GB
- Backups
- 7 days free retention
- Billing
- Monthly / hourly / yearly
Billing rules
| Item | Billing | Rules |
|---|---|---|
| Instance type | Hourly, monthly or yearly prepaid | HA edition is priced for two nodes; read replicas bill separately as single-node types |
| Storage | Per GB · month | RSSD cloud disk; online expansion bills at the new capacity |
| Backup storage | Per GB · month | Backup space equal to the instance storage is free; the excess bills by capacity |
| Public access | By elastic IP and bandwidth | Off by default; when enabled, bills under elastic IP and fixed bandwidth rules |
| Resizing | Price difference | Upgrades charge the difference; downgrades are prorated over the remaining term |
Unit prices vary by region; the pricing page and console show live quotes.
Managed or self-hosted
Only the dimensions that can be measured.
| Dimension | ZOSOC Cloud Database MySQL | Self-hosted MySQL on a cloud host | Major vendor managed MySQL |
|---|---|---|---|
| Time to provision | 3 to 5 minutes | Hours (install, tune, replicate) | Minutes |
| High availability | Automatic failover within 30 s, same address | Build your own MHA / Orchestrator | Automatic failover |
| Backup and restore | Daily full + binlog, restore to any second | Hand-written scripts, manual restores | Automatic backups, point-in-time recovery |
| Read/write splitting | Built-in proxy, no code changes | Application layer or your own ProxySQL | Usually a separately purchased proxy |
| Monitoring and slow log | 12 metrics, built-in slow query analysis | Your own Prometheus + exporter | Built in |
| Operations effort | 0 DBA inspections | Dedicated staff for inspections, patches, expansion | 0 DBA inspections |
| Connectivity with cloud hosts | Same VPC, 0.5 ms, whitelist control | Same VPC | Same VPC |
| Technical support | 24×7, 90-second response, incl. slow query and capacity advice | On your own | Tickets, billed by support tier |
Managed MySQL in three steps
Create the instance, whitelist your application and change the connection string.
Create the instance
Choose region, version, architecture, type and storage, in the same VPC as your application.
Create an account and whitelist
Create the application account and grant its database; add the application instances' private range to the whitelist.
Import data and switch the connection string
Import with mysqldump or migrate online via replication; point the application at the private address and enable the backup policy.
Frequently asked questions
The most common questions on versions, failover, backups and connections.
Existing connections drop at the moment of the switch; the address stays the same, so the application just reconnects. Enable automatic reconnection and short timeouts in your connection pool; failover usually completes within 30 seconds.
Small datasets import with mysqldump. For zero downtime, configure the managed instance as a replica of your self-hosted database for incremental sync, then stop writes, let it catch up and switch the connection string — the outage is usually minutes.
Daily full backups are kept 7 days by default and up to 30; binlogs are kept for the same window, so you can restore to any second inside it, into a new instance to switch to or export from.
Yes, inside the maintenance window with an automatic backup first; we recommend cloning an instance with point-in-time recovery to validate SQL compatibility on 8.0 beforehand.
Instances are reachable only inside the VPC by default; a public address can be enabled with a whitelist of sources. Use it only for migrations and temporary operations — production connections should stay on the private network.
Raise max_connections in the parameter group (bounded by memory) and use a connection pool on the application side; add read replicas for read traffic, and move up a type if the total still falls short.
Usually under a second within a region; monitoring exposes a replication-lag metric with alerts, and consistency-sensitive queries can add a SQL hint to force the primary.
Common parameters are changed through parameter groups; most apply online and a few need a restart. There is no operating-system or file-level access.
Use point-in-time recovery to restore the instance to the second before the drop as a new instance, export the table and import it back into production; writes continue throughout.
24×7 technical support with a 90-second response; slow query, connection and capacity planning questions can go straight to an architect for a 1-on-1 analysis.
Hand us the database,
you write the business logic
Provision the HA edition in 3 minutes, billed hourly. Send your current database size and QPS to an architect for a sizing recommendation.
- 30 sPrimary-standby failover
- 90 sTicket and live-chat response
- 1-on-1Slow query and capacity planning advice
- SLA credits99.95% availability commitment