Deployment

Agent runtime

Deploy the Dataclone agent where it fits your data path

The Dataclone agent can run on a bare-metal host, in a virtual machine or in a container. For maximum data-path performance, deploy the agent on bare metal and validate the complete host, storage and network path under the intended workload.

Control-plane runtime

Docker Compose for the control plane

The current Dataclone control plane is delivered as Docker Compose and can run on suitable customer-operated or Dexor-operated infrastructure by arrangement. The control plane is not on the payload path, so the agent’s bare-metal performance recommendation does not apply to it.

Agent operating systems

Linux today. Windows through WSL. Native Windows next.

The Dataclone agent runs directly on Linux and on Windows through Windows Subsystem for Linux. A native Windows agent is planned for a later release.

Deployment model

Files move directly between agents. You control each transfer.

Run the same Dataclone agent at each endpoint. Your files move directly between the agents over the connectivity you choose. Our separate on-premises control plane starts the job and reports status without relaying the payload.

Dataclone direct-path model

On-prem deployment

Dataclone direct-path architecture Two customer-controlled systems each contain one Dataclone software agent. A direct payload path joins the agents through customer-provided WAN connectivity. A separate on-premises control plane reaches the two agents through two independent control and status routes and does not relay payload. OPERATOR UI + API ON-PREM CONTROL PLANE Orchestration + observation control/status · not payload relay CONTROL / STATUS SITE A SITE B AGENT A AGENT B DIRECT AGENT-TO-AGENT PAYLOADRMDT / CUSTOMER-PROVIDED WAN Customer-controlled system Dataclone agent A inside Customer-controlled system Dataclone agent B inside Dataclone direct-path model Site A and Site B are arranged horizontally. Each customer-controlled system contains a Dataclone agent facing the direct payload path. Separate foreground control and status routes connect both agents to an on-prem control plane above. The opaque WAN cloud hides the payload while it crosses existing connectivity. SITE ACustomer-controlled system DatacloneAgent SITE BCustomer-controlled system DatacloneAgent DIRECT AGENT-TO-AGENT PATH ON-PREM CONTROL PLANEOperator / UI + APIInitiate · observe · retryOrchestrationControl and status onlyCONTROL / STATUS WAN Vertical mobile Dataclone direct-path modelSite A and Site B are arranged vertically. The direct payload path crosses an opaque WAN cloud beside the on-prem control plane. Separate control and status routes connect the control plane to both agents. SITE ACustomer-controlled systemDatacloneAgent DatacloneAgentSITE BCustomer-controlled system DIRECT PATHAgent-to-agent ON-PREM CONTROLOperator / UI + APIStart · monitor · retryOrchestrationControl / status onlyCONTROL / STATUS WAN

Operator UI · API-assisted workflow

Control plane: orchestration + observation · Control/status, not payload relay

Control/status paths to both agents

Site A

Customer-controlled system

Dataclone agent

Agent A

Direct agent-to-agent payload path

Site B

Customer-controlled system

Dataclone agent

Agent B

Your connectivity. RMDT runs over existing IP infrastructure without proprietary network hardware. You choose the interface for the payload path.

Control-plane hosting. The diagram shows the customer-hosted option. Dataclone can also use a control plane hosted by Dexor by arrangement. In that model, the agents connect to the hosted controller while file payload still moves directly between the participating agents.

Explore Dataclone managed file transfer

Read the architecture description

RMDT transport. We built it for high-capacity WAN paths affected by latency, jitter and packet loss. Route, hosts, storage, workload and configuration still shape the result.

How we test it

We agree the endpoints, route, workload and useful criteria before the run. When your current method gives us a meaningful baseline, we compare against it on the same path.

Your two sites run the same Dataclone agent software. Both agents connect to our separate on-premises control plane for job orchestration and status. The file payload moves directly between the agents over your connectivity; the control plane does not relay it. You operate the transfer through the Dataclone UI or an API workflow. Deployment and path prerequisites are reviewed during the technical discussion.

Control-plane hosting

Choose who hosts the control plane

Run it in your environment

The control plane is delivered as Docker Compose and can run on suitable infrastructure you operate. Agents connect to that controller for organization, workspace, permission, orchestration and status functions.

Use a controller hosted by Dexor by arrangement

Dexor can host, operate and update the control plane by arrangement while the agents remain on the systems that hold your data. The hosted controller isolates organizations and handles transfer-control data including file size. Agents connect to that controller, but file payload continues to move directly between the participating agents. Dexor provides agent software updates.

For the hosted option, Dexor operates and updates the control plane.

In either model, agents connect to the selected controller for control and status. The controller does not relay the file payload.

Network requirements

Permit the control channel and selected data ports

Agents establish HTTPS/WSS control connections to the selected controller. RMDT and RSDT data traffic use configured UDP ports. DMTP data traffic uses configured TCP ports. Customers permit the selected data-port range between participating agents according to sender/receiver roles.

Control and data paths

You choose the path. We keep control separate.

Dataclone runs on your systems. The agents move the payload directly. Our control plane coordinates the job and reports status without relaying the file.

Review the 2024 FTP-over-TCP lab conditions when defining a route- and workload-specific validation.