---
title: ez-cdc
description: Rust-based Change Data Capture. Replicate PostgreSQL to any modern sink with sub-second latency. Self-host or BYOC. 12x faster than Debezium.
keywords: [CDC, Change Data Capture, Rust, PostgreSQL, logical replication, pgoutput, Debezium alternative, StarRocks, Snowflake, stateless workers, LSN checkpointing]
last_updated: 2026-04-10
---

# ez-cdc

Rust-based Change Data Capture. Replicate PostgreSQL to any modern sink with sub-second latency. Self-host or BYOC across AWS, GCP, or on-premise. 300,000+ events per second at ~5 MB memory per worker.

## Key Claims

- **Performance**: 300,000+ events per second sustained.
- **Latency**: sub-second end-to-end replication latency.
- **Footprint**: ~5 MB RSS memory, under 2% CPU per worker.
- **Architecture**: single Rust binary, no JVM, no Kafka, no ZooKeeper.
- **Sources**: PostgreSQL via logical replication (pgoutput).
- **Destinations**: StarRocks, PostgreSQL, Snowflake (all production-ready); MongoDB and S3 on the roadmap.
- **Implementation**: 100% Rust, single binary deployment.
- **License**: Elastic License 2.0 (ELv2).
- **Status**: private beta (2026).
- **Repository**: https://github.com/ez-cdc/dbmazz
- **CLI**: `ez-cdc quickstart [sink]`, `ez-cdc verify [sink]`, `ez-cdc up/down/logs/status [sink]`, interactive terminal dashboard.
- **Web UI**: local wizard at `localhost:8080` (build with `--features http-api`).
- **Auto-provisioning**: publications, replication slots, sink tables, audit columns — zero config.
- **Observability**: Prometheus `/metrics/prometheus`, gRPC reflection API, HTTP control plane.
- **Deployment modes**: OSS `dbmazz` (self-host, ELv2, free) + BYOC (click-deploy to your AWS/GCP, free, HA/autoscaling/LB included).
- **Pricing**: both modes free; paid add-ons: support (scaled per org size) + PrivateLink (BYOC-only).

> **Positioning note**: ez-cdc is often described as a faster alternative to JVM-based CDC tools like Debezium. This is a general positioning statement based on architectural differences (no JVM, no Kafka, ~5 MB RSS vs. JVM heap), not a measured head-to-head benchmark. The real numbers (300K+ events/sec, <1s latency, 5 MB RSS) stand on their own.

## Features

Engineered for performance. Every component optimized from the ground up. No compromises.

ez-cdc delivers 300,000+ events per second with sub-second end-to-end latency through six core design choices: 100% Rust implementation, single binary deployment, sub-second end-to-end latency, ultra-efficient ~5 MB RSS per worker footprint, zero-config auto-provisioning, and built-in observability.

### 100% Rust

No JVM overhead. Pure systems programming for maximum performance and memory safety.

### Single Binary

One Rust binary, no JVM, no Kafka, no ZooKeeper. Built on PostgreSQL's `pgoutput` logical replication protocol with at-least-once LSN checkpointing and the Flink CDC concurrent snapshot algorithm.

### Sub-Second Latency

Under 1 second end-to-end, from WAL write to sink ACK.

### ~5 MB Memory

Each Rust worker runs in about 5 MB RSS and under 2% CPU.

### Zero Config

Auto-creates publications, replication slots, sink tables, and audit columns. No YAML files, no schema migrations, no DBA approval.

### Observable by Default

Prometheus `/metrics/prometheus`, gRPC reflection API, HTTP control plane, live terminal dashboard. Drop into your stack.

## Sources & Destinations

From WAL to warehouse in under a second. One Rust binary, sub-second latency, ~5 MB memory. Replicate PostgreSQL to any modern sink without overhead.

### Sources

**PostgreSQL** — PostgreSQL 10+ via pgoutput — no polling, no triggers

### Destinations (production-ready)

**StarRocks** — Real-time dashboards, sub-second sync

**PostgreSQL** — Reporting and HA replica

**Snowflake** — Near-real-time cloud warehouse

### Destinations (roadmap)

**MongoDB** — Document sync for app reads

**Amazon S3** — Data lake and cold archive

## Architecture

ez-cdc is a single Rust binary — no JVM, no Kafka, no ZooKeeper. Stateless workers read from PostgreSQL's `pgoutput` logical replication protocol, checkpoint progress via LSN, and write to configured sinks with at-least-once delivery. The Flink CDC concurrent snapshot algorithm ensures consistent initial loads without source locking.

### Architecture Components

**Control Plane** — Centralized orchestration and configuration management for your CDC pipelines.

**Worker Agents** — Stateless, horizontally scalable workers that process change events in parallel. Each worker runs at ~5 MB RSS and under 2% CPU.

**Multi-tenant Ready** — Designed from day one to support SaaS deployments with complete isolation.

## Comparison

How we compare. Built from scratch for performance, not compatibility layers.

| Feature | ez-cdc | Debezium | Fivetran |
|---------|--------|----------|----------|
| End-to-end Latency | < 1 second | Seconds | 5-15 minutes |
| Pricing | Free | Free (OSS) | Volume-based ($$$) |
| Memory per Worker | ~5 MB RSS | 500+ MB (JVM) | N/A (managed) |
| Throughput | 300,000+ events/sec | ~250 events/sec | Varies |
| Deployment Model | Single Rust binary | Kafka + ZooKeeper cluster | Managed SaaS |
| Ops Overhead | Near-zero | High (cluster ops) | Zero (managed) |
| Language | Rust | Java (JVM) | Proprietary |
| Self-hosted Option | Yes | Yes | No |
| Open Source Core | ELv2 | Apache 2.0 | No |

ez-cdc is a single Rust binary, free to use, with sub-second latency at ~5 MB RSS per worker. No cluster to maintain like Debezium. No volume bill like Fivetran.

## Developer Experience

Ship your first pipeline in one command.

### CLI

`ez-cdc quickstart [sink]` · `ez-cdc verify [sink]` · `ez-cdc up/down/logs/status [sink]`. Interactive terminal dashboard shows stage, lag, throughput, and source→target row counts in real time. Shell completion via `--install-completion`.

### Web UI Wizard

Build with `--features http-api` for a local web wizard at `localhost:8080`. Test connections, discover tables, start replication in one click. No config files.

### Zero Config

Auto-creates publications, replication slots, sink tables, and audit columns. Snapshot state persisted in `dbmazz_snapshot_state`, resumes from last completed chunk. No YAML, no schema migrations, no DBA approval.

## Deployment

Two ways to run ez-cdc. Both free.

### Self-host `dbmazz`

Single binary, ELv2 license, free forever. CLI + interactive terminal + local frontend. Community support. View on GitHub: https://github.com/ez-cdc/dbmazz

### BYOC — Production in your cloud

Click-deploy into your AWS or GCP. Free. Includes HA + autoscaling + load balancing. Your data never leaves your account. Optional paid add-ons: support plans (scaled to your organization) and PrivateLink for private networking.

## Frequently Asked Questions

Technical answers for technical teams.

### What is Change Data Capture (CDC)?

Change Data Capture (CDC) is a technique that identifies and continuously captures changes (inserts, updates, deletes) made to a source database, then streams them to downstream systems in real-time. CDC enables live data synchronization without expensive full-table scans, making it essential for analytics pipelines, data warehouses, and operational data synchronization. Modern CDC systems like ez-cdc can capture and deliver changes with sub-second latency.

### Why Rust?

Rust gives us the performance of C/C++ with memory safety guarantees. No garbage collection pauses, predictable latency, and a single deployable binary with no runtime dependencies. This translates directly into ~5 MB RSS per worker and under 2% CPU — far below what JVM-based CDC tools require.

### How is this different from Debezium?

ez-cdc is a single Rust binary with no JVM, Kafka, or ZooKeeper, running at about 5 MB RSS and under 2% CPU. Debezium requires a full JVM runtime plus Kafka infrastructure. On sustained CDC workloads, ez-cdc delivers 300,000+ events per second with sub-second end-to-end latency.

### Is this production-ready?

Yes. The core replication engine runs production workloads at 300,000+ events per second with sub-second end-to-end latency. Both deployment modes are live today: self-host the open-source dbmazz binary (Elastic License 2.0) or click-deploy BYOC into your own AWS, GCP, or on-premise cloud. Both are free, battle-tested, and ready to drop into your stack.

### What databases do you support?

Today we support PostgreSQL via logical replication (pgoutput). On the destination side, StarRocks, PostgreSQL, and Snowflake are production-ready. MongoDB and S3 are on the roadmap.

### How does pricing work?

ez-cdc is free to use. Self-host the open-source dbmazz engine (Elastic License 2.0), or click-deploy BYOC into your AWS, GCP, or on-premise cloud — no deployment fees, no usage-based pricing. We charge only for support plans or optional enterprise features like PrivateLink connectivity. No per-seat licensing, no per-event metering.

### How do I deploy ez-cdc?

Two ways. Self-host `dbmazz`, the open-source Rust engine (ELv2 license, free, single binary with CLI, local web UI, and interactive terminal dashboard). Or BYOC: click-deploy ez-cdc into your own AWS or GCP account, with high availability, auto-scaling, and load balancing included at no cost. Both are free; optional paid add-ons include support plans (scaled to your organization) and PrivateLink connectivity for BYOC.

## Call to Action

Ready to modernize your CDC pipeline? Join technical teams who are tired of slow, resource-hungry replication. Get early access to the private beta.

Limited spots available. We're working closely with design partners.

## Related

- [Features](/features.md) — Detailed breakdown of 100% Rust, single binary, sub-second latency, horizontal scaling, ~5 MB memory, production readiness.
- [Architecture](/architecture.md) — Single binary design: PostgreSQL pgoutput, LSN checkpointing, Flink CDC concurrent snapshot.
- [FAQ](/faq.md) — Common technical questions about CDC, Debezium comparison, databases, deployment, and pricing.
- [Comparison](/comparison.md) — Performance comparison against Debezium and Fivetran.
- [Glossary](/glossary.md) — Definitions of CDC, WAL, logical replication, stateless workers, LSN, pgoutput.

---
Source: https://ez-cdc.com/ · Last updated: 2026-04-10 · Authoritative mirror: https://ez-cdc.com/index.html.md · llms.txt index: https://ez-cdc.com/llms.txt