Engineered for scale, built to be auditable
Bridge is a distributed, container-based integration engine. The same build runs on a laptop, a single VM, or a multi-node cluster — and the same engine powers a 10-billion-row daily pipeline at a public-safety agency.
Capture from your database logs
Bridge captures change data by parsing the source database's native transaction logs — Oracle Redo, MySQL binlog, PostgreSQL WAL, Kingbase WAL, and DM Redo — reading the source transaction logs directly.
The method is covered by a granted patent (a non-intrusive high-speed incremental-extraction technique). Your DBA approves a read-only log reader; nothing else touches the database.
- ✓Reads the database transaction log directly
- ✓Read-only access, no triggers, no shadow tables
- ✓Sub-minute CDC lag in typical setups (P95 ~40–60s)
POSTGRES · parsing WAL · slot=sumeta_cdcLSN 4/F8A2C1B0 INSERT orders +1LSN 4/F8A2C1C8 UPDATE orders +1LSN 4/F8A2C2D0 DELETE sessions +1# committed to Snowflake · offset saved · resume-ready
Numbers from real deployments
Not marketing math — these are figures from production engagements.
| Metric | Figure | Context |
|---|---|---|
| Single-node throughput | 120 MB/s | 8-core / 32 GB node, bulk load |
| High-scale capture | 280k rows/s | Aerospace research institute (real deployment) |
| Daily volume (single client) | 10B+ rows/day | Public-safety agency telemetry |
| File push | 60 MB/s | Sustained, compressed transfer |
Throughput and latency depend on hardware, network, and workload shape. Run our open benchmark kit against your own sources before you sign anything.
It has to keep running
Checkpoint & resume
Every sync stores an offset. A crash or network blip resumes from the last committed point — no re-snapshot, no duplicates.
Auto-rollback
Failed batches are rolled back atomically. The target is never left in a half-applied state.
Cluster, no SPOF
Multi-instance deployment with no single point of failure. Scale horizontally and linearly.
Dual channel scheduling
Mix batch and real-time jobs on shared workers — e.g. 16 channels running 16 batch or 64 real-time jobs concurrently.
Run it where your data lives
Single VM / container
Minimum 4-core / 8 GB. Perfect for evaluation and mid-size teams.
VPC / Kubernetes
Recommended 8-core / 32 GB per node. Horizontal scale, no single point of failure.
On-prem / isolated
Sync across security gateways via file exchange. No inbound connectivity required — built for regulated environments.
Put the engine to the test
Download a free evaluation license and run the benchmark kit against your own sources.