Compare

How Remac stacks up.

An honest, technical comparison on deployment, operations, cost, recovery, and architecture. Not a feature checklist.

Remac vs Debezium

JVM ecosystem
Infrastructure
Remac
Single binary, zero external dependencies. One process, one config file.
Debezium
Multiple deployment modes, all JVM-based. Kafka Connect mode needs a Kafka cluster, Connect workers, and JVM tuning. Every mode requires JVM operations and configuration.
Destinations
Remac
Writes directly to Kafka, S3, PostgreSQL, RabbitMQ, files, HTTP webhooks, and more with native multi-sink fan-out.
Debezium
No native multi-sink fan-out. Reaching multiple destinations requires either a full Kafka cluster as an intermediary or running separate instances per sink.
Resource usage
Remac
Compiled Go binary with low memory usage. Runs comfortably on a 4 vCPU instance. No JVM tuning.
Debezium
JVM-based. Production heap typically 1-4 GB depending on schema size. GC pauses affect tail latency.
Recovery
Remac
Built-in physical + logical point-in-time recovery and REDO. Automated recovery from any failure point.
Debezium
Relies on Kafka retention for replay. Manual intervention when offsets drift or connectors fail.
Operational cost
Remac
One binary to deploy, one process to monitor. No dedicated CDC engineer required. Fraction of the operational budget.
Debezium
JVM tuning, monitoring, and infrastructure management regardless of deployment mode. Kafka Connect mode adds a full Kafka cluster to the operational surface.

Remac vs Fivetran

per-row SaaS
Pricing model
Remac
Flat licensing. Predictable costs that don't penalize success or data volume.
Fivetran
Usage-based pricing (MAR). Every unique row that changes in a month counts toward your bill. Costs scale with your data volume.
Latency
Remac
Real-time streaming from the transaction log. Low-millisecond delivery to every sink.
Fivetran
Batch-oriented syncs on minute-level intervals. Not real-time, regardless of what the marketing says.
Control
Remac
Runs in your environment, your VPC, your rules. Data never leaves your infrastructure.
Fivetran
SaaS-managed. Limited control over infrastructure, routing, or processing logic. Data transits through their servers by default.
Recovery
Remac
Built-in PITR (physical + logical). Restore to any point in time without external help.
Fivetran
No PITR. Recovery depends on Fivetran's internal retry logic and support response times. You don't control the recovery path.
Operational cost
Remac
Self-hosted, no metered fees. Cost is the compute you already control.
Fivetran
Usage-based pricing becomes significant at scale. Enterprise tiers add more. Your Fivetran bill grows every time your product succeeds.

Remac vs Airbyte

batch ELT platform
Focus
Remac
Purpose-built for transaction log processing. Does one thing, does it well.
Airbyte
A broad ELT platform with 600+ connectors, built for batch data warehousing. CDC is a feature, not the architecture.
Complexity
Remac
Single binary. No scheduler, no orchestrator, no metadata database.
Airbyte
Scheduler, workers, orchestrator, metadata database, web UI. Many moving parts that all need to stay alive.
Latency
Remac
Streams changes as they happen from the WAL. Real-time, not batch.
Airbyte
Syncs in configurable batches. Real-time is not the design goal and the architecture doesn't support it.
Recovery
Remac
Physical + logical PITR and REDO. Resume from any point without full re-sync.
Airbyte
No built-in PITR. Recovery means re-syncing from scratch.
Operational cost
Remac
Minimal infrastructure footprint. One process on modest compute. No per-row charges.
Airbyte
Self-hosted Airbyte needs substantial infrastructure. Cloud Airbyte uses usage-based pricing that scales with data volume.

Remac vs AWS DMS

cloud-locked
Portability
Remac
Runs anywhere: AWS, GCP, Azure, bare metal. Your infrastructure, your choice.
AWS DMS
The DMS engine runs only in AWS. Your CDC pipeline logic, monitoring, and operations are locked to one cloud provider.
Configurability
Remac
Fully configurable pipeline with middlewares (filtering, transformation, etc). When something needs tuning, you tune it.
AWS DMS
Configuration through JSON-based task definitions in the AWS console. When things go wrong, debugging is opaque.
Multi-target
Remac
Native multi-sink fan-out from a single source stream. One replication slot, many destinations.
AWS DMS
One target per migration task. Fan-out means duplicating the entire task.
Recovery
Remac
Full physical + logical PITR. Restore to any moment in the recovery window.
AWS DMS
Limited restart capability. No point-in-time recovery. No REDO.
Operational cost
Remac
Transparent resource usage on compute you control. No surprise charges, no metered overages.
AWS DMS
DMS instance hours + data transfer + CloudWatch. Costs are opaque until the bill arrives.

Remac vs pg_logical / native replication

postgres-to-postgres only
Destination support
Remac
Kafka, S3, PostgreSQL, RabbitMQ, files, HTTP webhooks. One source, many destinations.
pg_logical / native replication
PostgreSQL-to-PostgreSQL only. No Kafka, no S3, no message queues. Fine until you need data anywhere else.
Fan-out
Remac
One-to-many fan-out from a single replication slot. No source overhead per destination.
pg_logical / native replication
One-to-one replication. Each additional destination requires a separate replication slot on the source.
Transforms
Remac
Configurable pipeline with middlewares (filtering, transformation, etc). Deliver only the changes that matter to each sink.
pg_logical / native replication
No transformation, no routing, no event-level processing. Publication-level row filters exist (PG 15+) but that's it. What the WAL produces is mostly what you get.
Recovery
Remac
Full physical + logical PITR that works across all sink types, not just PostgreSQL.
pg_logical / native replication
Limited to what PostgreSQL provides natively. No cross-engine PITR.
Want the numbers?

See our CDC benchmarks on real AWS infrastructure.

View performance benchmarks →

Deploy once. Forget the pipeline exists.

Your data arrives where it needs to be, in order, with delivery guarantees — and your team never thinks about it again.

Read the docs