The point-to-point migration method uses a migration agent to transfer the entire source end to the target machine by directly transmitting source-end data to the target machine. This method eliminates the need to back up the source-end system—while the source-end data is being transmitted to the HCI, the platform starts the target-end virtual machine to receive the migrated data. It supports both scheduled incremental data synchronization and manual incremental data synchronization, ensuring that the business cutover window can be controlled manually.

Applicable Scenarios

This migration method is recommended for the vast majority of business scenarios, with service downtime of approximately 5 to 10 minutes during the business cutover. The specific downtime is related to the startup time of the business services.

Migration time is relatively controllable. Data transmission can start as early as possible, with the final data synchronization and service switchover performed after the full data transmission is completed and during the business downtime window.
Point-to-point migration does not require backing up data from the source VM to the SCMT, which reduces the storage requirements of the SCMT. Data is transmitted directly to the target side without passing through the migration platform, shortening the overall data migration time.

Migration Process

The migration process for Agent-based Point-to-point Migration mode in Sangfor Cloud Migration Platform is as follows:

A diagram of a software system

AI-generated content may be incorrect.

The source machine installs the migration agent, and the migration tool has two modes: active and passive.

• In the active mode, the migration agent at the source actively connects to the SCMT and receives tasks issued by the SCMT.

• In the passive mode, SCMT will take the initiative to connect the source achine migration agent to issue migration tasks.

SCMT calls HCI's API interface to create the destination virtual machine and automatically loads the bare-metal restore to start.
After modifying the bare metal restore IP address, the destination virtual machine actively connects to the SCMT, obtains the migration task through the SCMT, and receives the data transmitted from the source machine migration agent.
Full data synchronization.
Incremental data synchronization.
Stop the source production, and perform production switching after synchronizing the last difference data.
The destination virtual machine starts compatibility processing and automatically installs performance optimization tools(vmTools).
The destination virtual machine restarts and takes over the production.

1.Since data is not retained on the SCMT with this migration method, verification cannot be performed during migration process. Verification can only be conducted on the target machine after the migration is complete.

2.When migrating a source VM with dynamically allocated disks, If the data volume exceeds 16TB (exceeding the size limit of dynamically allocated disks), the allocation will automatically switch to thin provisioning.

3.When switching to HCI, simply stop the business services on the source side—do not shut down the source-side operating system.

4.Migrating data from a single disk on the source side to multiple disks on the target side is not supported.

5.Migration of external storage with mounted partitions is supported; migration of raw disks without file system partitions is not supported.

6.Migration of file storage mounted on the source VM is not supported.

7.Migration can only be initiated after installing the agent on source VM and preparing the target VM. HCI versions 6.0.0 R5, 6.3.0, and later support automatic creation of the target VM after the HCI and SCMT are already integrated.

8.For agent-based to agentless migration on ARM (where the target machine is booted from PE boot media), the target machine must have a minimum of 3GB of memory.

9.The maximum number of disks supported is 64 for HCI 6.11.1, and 16 for HCI versions prior to 6.11.1.

Migration Time

The initial synchronization only transfers the actual used data blocks within the virtual machine's disk, not the entire provisioned capacity. For instance, if a VM has a provisioned disk of 100 GB but only 50 GB is utilized, the initial sync will transfer approximately 50 GB of data. Subsequent replications are incremental, capturing only the data blocks that have changed since the last sync. Following the final incremental sync, a planned service cutover is required. This involves stopping the services on the source machine and redirecting operations to the target. For point-to-point migrations, the service interruption during this cutover window is typically between 5 to 10 minutes.

Phase	Migration Phase	Tasks	Duration / Impact
1	Pre-Migration Environment Setup	1.1 Deploy SCMT Server (One-time setup) (10 min) 1.2 Install Agent on all source machines (2 min/machine) 1.3 Create Virtual Machines on target (2 min/machine)	Duration: around 20 minutes Impact: no impact on source services
2	Data Replication	2.1 Full Data Replication 2.2 Multiple Incremental Replications (Automatic or manual)	Duration: Total Data Volume / Migration Speed - Max speed: 100 MB/s (1 GbE), 500 MB/s (10 GbE) Source Resource Utilization: - Agent CPU usage: < 3% - Agent Memory peak: < 260 MB - Recommended available memory on source: > 1 GB - Default source storage performance cap: 30% (Queue Depth: 4) - Adjustable - Default source network bandwidth cap: 300 Mbps - Adjustable
3	Service Cutover	3.1 Manually stop services on source machine (2 min) 3.2 Final data sync & execute cutover (1-2 min) 3.3 Restart target VM (Auto compatibility & VMTools injection) (2 min) 3.4 Modify service network (As required) (approximately 2 min) 3.5 Services resume, hosted by HCI (2-5 min)	Duration: 5-10 minutes Impact: Source services stopped

Resumable transfer

• If the network interruption between the source and target ends lasts less than 15 minutes, the migration task will not be affected. After the network resumes normal operation, incremental data will be automatically resumed for transmission.

• If the network interruption between the source and target ends lasts more than 15 minutes, data synchronization will fail. After the network resumes normal operation, incremental data transmission can be resumed either manually or automatically (by waiting for the next synchronization time point).

• If the source/target end is shut down and restarted or disconnected and reconnected, data synchronization will fail, but there will be no impact on the source-end business. After recovery, full data transmission can be resumed either manually or automatically (by waiting for the next synchronization time point).

• If the server-side is restarted or shut down during the migration process, data synchronization will fail, but there will be no impact on the source-end business. After recovery, full data transmission can be resumed either manually or automatically (by waiting for the next synchronization time point).

{{ $t('index.defaultHeader.logo') }}

Sangfor Cloud Platform (SCP)

Point-to-point Migration (agent-based)