Platform
Data migration tool with checkpoints and CDC
A data migration tool that moves large tables in batches you can resume, then keeps the target in sync until cut-over.
18 min
application downtime at cut-over
Run timeline, ERP-Oracle to ERP-PG
- 1Discovery6m
- 2Compatibility & Mapping9m
- 3Risk & Approval (approved in advance)0m
- 4Strategy2m
- 5Sandbox Dry Run21m
- 6Schema Migration32m
- 7Data Migration13h 48m
- 8Validation42m
- batches
- 42,838
- average
- 43,100 rows/s
- peak
- 61,200 rows/s
- resumes after network interruptions
- 2
Coverage
What does the data migration cover?
One run moves the schema, the data and the source changes made until cut-over.
Schema and code
Tables, constraints, sequences and converted code, deployed to the target before any data moves.
Large tables
Fixed-size batches, several tables in parallel. Every completed batch is a checkpoint.
Changes during the copy
CDC picks up changes made on the source, so the source stays in use.
Snowflake and Databricks too
CDC keeps the target in sync on PostgreSQL, Snowflake and Databricks targets.
How it works
How does the data migration work?
Schema first, then data, then keep in sync until you choose to cut over.
1. Schema first
Nothing is copied until the schema is on the target.
- 5 schemas
- 1,036 tables
2. Data in checkpointed batches
If the connection drops, the run resumes from the last checkpoint instead of starting the table again.
- 2,141,876,402 rows
3. Keep in sync with CDC
After the full copy, CDC applies source changes. You choose when to cut over, with the lag and validation verdict in front of you.
- 1,204,318 changes synced
- Lag 4 s
4. Cut over when ready
Stop writes on the source, let the last changes apply and switch the application. The Migration Report records the lag and the downtime.
- Downtime 18 min
Report
What does the Migration Report record?
From the example run, ERP-Oracle (Oracle Database 19c) to ERP-PG (PostgreSQL 16.4):
| Measure | Example result |
|---|---|
| Rows copied | 2,141,876,402 rows, 412 GB, 1,036 tables |
| Data copy | 23:10 (2 Oct) to 12:58 (3 Oct) UTC, 13h 48m |
| Batching | 50,000-row batches, 8 tables in parallel, 42,838 batches |
| Throughput | average 43,100 rows/s, peak 61,200 rows/s |
| Interruptions | 2 resumes after network interruptions |
| CDC | 1,204,318 changes applied from 3 Oct 13:40 UTC to cut-over |
| Cut-over | 4 Oct 2026 02:00 UTC, lag 4 s, application downtime 18 min |
| Total run | 15h 40m including schema and validation |
- Changes applied
- 1,204,318
- Lag at cut-over
- 4 s
Related
Related features
- Data migration validation17 checks per table, including row counts and checksums.
- Stored procedure conversionProcedures, functions, triggers and views, compiled before deploy.
- Cloud or self-hosted deploymentRun it in Datalobe Cloud or self-hosted.
Questions
Frequently asked questions
Can you migrate without any downtime?
Downtime is minimal, not none. It is limited to the cut-over itself. In the example run, application downtime was 18 min.
What happens if the connection drops during the copy?
The run resumes from the last completed batch. In the example run, it resumed twice. In a rehearsal of the example migration, an interrupted batch was not resumed, and validation caught 1,250 missing rows.
Do we have to stop using the source database during the copy?
No. The source stays in use. CDC applies the changes made during and after the copy until you cut over.
How do we know the data arrived intact?
Validation runs after the copy: 17 checks per table, including row counts and checksums on all rows. See data migration validation.
Free assessment
Plan your cut-over with real numbers
Start with an Assessment Report on your own database.