How YY Scaled Its Database Platform: From Manual Ops to Intelligent Automation

1. YY Database Team Challenges

Early issues included frequent quality incidents, slow resource‑request response, reactive security firefighting, and high cost due to low resource utilization.

Quality accidents harmed service reputation.

Resource demand response lagged behind business needs.

Security incidents were often handled post‑mortem.

Low resource utilization drove high costs.

2. Database Platform Goals

The platform aims for one‑stop service, minute‑level deployment, automated change operations, minute‑level monitoring and alerting, rapid fault diagnosis, auditability, measurable cost, and traceable security changes.

3. Resource Pool Management

YY operates multiple IDC and public‑cloud data centers with diverse storage (SAS, SATA, SSD, PCIe) and supports multiple MySQL, MongoDB, and Redis versions, allowing multi‑instance deployment on a single machine.

Business users select database type, package size, and configuration from a resource‑pool interface, enabling minute‑level interactions.

4. High‑Availability Proxy Architecture

YY uses a high‑availability proxy layer (OSPF/VIP) that routes writes to a primary IDC while reads can be served locally, providing load balancing, read/write separation, and special handling for large SQL queries.

Multi‑tenant design allows new services to be configured via the GK system without full cluster deployment.

5. MyShard – Multi‑Write Solution

MyShard implements eventual consistency by adding version numbers to tables, enabling multi‑write across regions and handling conflicts by preferring the highest version.

6. Database Synchronization Services

Provides tools to sync selected tables or fields between databases, support for heterogeneous targets, performance‑optimized sync, cache updates, RabbitMQ notifications, and cross‑database replication.

7. Database Quality Platform

Features rapid alarm detection, instance‑level and business‑level diagnostics, event tracking, and a global dashboard showing alerts, warnings, and health metrics via Kafka‑fed data streams.

8. Cost Management

Cost is quantified per instance, business, and organization, with dashboards showing usage and spending, and one‑click tier‑up/down of resource packages while ensuring data consistency during migrations.

9. Security Measures

Includes baseline network and OS security, cross‑region backup (primary and secondary), data‑consistency checks with automated repair, controlled isolation and decommission procedures, and SQL execution safeguards that rewrite risky statements (e.g., DROP to RENAME).

10. Future Outlook

YY plans to advance toward intelligent automation for scaling, monitoring, and capacity planning, enhance SLA transparency, strengthen SQL audit capabilities, and further improve resource utilization to lower operational costs.