PgDog offers a transparent routing layer for PostgreSQL, enabling horizontal scaling without application code changes. It uses logical replication for zero-downtime sharding and intelligently routes queries to achieve NoSQL-like performance while preserving SQL capabilities.
Funding
Funding not disclosed
Founders
Product
Problem
Scaling PostgreSQL databases to handle growing data volumes and user loads often requires complex sharding strategies and significant application re-architecture. Traditional approaches can lead to downtime during migration or require extensive modifications to existing application codebases.
Solution
PgDog provides a transparent routing layer for PostgreSQL, enabling horizontal scaling without application code changes. It leverages logical replication to shard data across multiple PostgreSQL instances with zero downtime. The system intelligently routes query traffic based on sharding keys, or transparently aggregates results from multiple shards for cross-shard queries. This architecture allows PostgreSQL to achieve NoSQL-like performance characteristics while retaining its robust SQL capabilities and data integrity.
Target Audience
Database administrators and developers managing large PostgreSQL deployments who require enhanced scalability and performance without extensive application refactoring.
Features
- **External Sharding Layer:** Operates independently of PostgreSQL, compatible with managed cloud services like AWS RDS and Google Cloud SQL.
- **Zero-Downtime Sharding:** Utilizes PostgreSQL's logical replication for seamless data migration and sharding without service interruption.
- **Intelligent Query Routing:** Parses SQL queries to direct traffic to appropriate shards based on defined sharding keys.
- **Cross-Shard Query Aggregation:** Automatically routes queries without sharding hints to all relevant shards and consolidates results transparently.
- **Parallelized Operations:** Employs a share-nothing architecture where shards function as independent databases, enabling parallel processing.
- **SQL-Compatible Performance:** Delivers high throughput and low latency, comparable to key-value stores, by scaling horizontally with added data nodes.
- **DDL Statement Distribution:** Distributes DDL statements (e.g., `CREATE TABLE`) across all shards to maintain schema consistency.
- **Limited Aggregate Function Support:** Supports `min()`, `max()`, `count()`, and `sum()` for cross-shard queries, with ongoing development for broader support.
- **Load Balancing:** Distributes read queries across PostgreSQL replicas using configurable strategies such as random, least active connections, and round robin.
- **Read/Write Splitting:** Automatically directs write operations to the primary instance while routing read operations to replicas.