Data flow

This page traces the exact path that data takes during a replication cycle, from the source VM's disk blocks to a committed recovery point at the target.

Replication cycle data flow

A single replication cycle for a VMware-to-AWS example proceeds as follows:

1. Control plane schedules the job

   The scheduler determines that a plan's RPO interval has elapsed and enqueues a replication job assigned to the Replication Appliance registered for the source site.

2. Replication Appliance picks up the job

   The agent on the Replication Appliance polls the job queue, receives the job, and begins execution. It reports Running status back to the control plane.

3. Snapshot creation

   The agent calls the vSphere API to create a snapshot of each protected VM. If application consistency is configured, the agent triggers VSS quiescence (Windows) or pre-freeze scripts (Linux) before snapshot creation.

4. CBT query

   For each disk, the agent queries the vSphere CBT API for the list of block ranges that changed since the previous cycle's snapshot ID. This produces a sparse list of dirty byte ranges.

5. Block read via VADP

   The agent reads only the changed blocks from the snapshot through the vSphere NFC (Network File Copy) API. Blocks are read in parallel across disks to maximize throughput.

6. Compress and encrypt

   Each block is compressed (LZ4 or ZSTD) and then encrypted with AES-256-GCM using the plan's data encryption key. This happens on-appliance before the block leaves the source network.

7. Transfer to Cloud Connector

   Encrypted blocks are streamed over TLS 1.3 to the Cloud Connector in the target AWS VPC. The Cloud Connector acknowledges each batch and writes it to the staging EBS volume.

8. Shadow disk commit

   After all changed blocks are received, the Cloud Connector patches the shadow EBS volume with the new data, producing a consistent point-in-time disk image that matches the source as of the snapshot timestamp.

9. Recovery point record

   The Cloud Connector creates an EBS snapshot of the updated shadow volume. The snapshot ID and its corresponding source timestamp are written to the recovery point index in the control plane database.

10. Source snapshot deletion

    The Replication Appliance deletes the vSphere snapshot created in step 3. The VM resumes normal I/O. The agent reports Completed status with transfer statistics.

What data crosses which boundary

In the SaaS model, only control-plane metadata (job status, recovery point index, plan definitions) transits to Datamotive. In self-hosted and sovereign deployments, nothing leaves the customer boundary.

Related docs

• Logical architecture

• Block transfer

• Encryption and compression

• Control plane