Estimate migration time and test transfer speed - Server Migration Center

The total time depends on the number of servers to migrate and the total data volume. Perform a test migration to get an accurate estimate. This topic describes how to estimate the time for a full or incremental migration and how to test the data transfer speed.

Background information

During a Server Migration Center (SMC) migration, the service first provisions a temporary intermediate instance in your Alibaba Cloud account. If you are migrating to a specific destination instance, that instance is used as the intermediate instance. Data is then transferred from your source server to a cloud disk on the intermediate instance. Finally, SMC creates an ECS image from the data on the cloud disk. For more information, see How SMC works. Therefore, the total migration time is the sum of the data transfer time and the image creation time. For details, see Estimate migration time.

Throughout the migration process, the transfer speed from the source server to the intermediate instance is the primary factor determining the migration time. For information about how to test the transfer speed, see Test the data transfer speed.

The following table describes the possible causes and solutions if the actual data transfer speed is lower than the tested speed.

Possible cause	Solution
The source server and the intermediate instance are in different regions or countries. Cross-region or cross-country data transfers can be slower than transfers within the same region.	If the issue is cross-region transfer, you can do one of the following: Migrate the source server to the same Alibaba Cloud region, generate an image, and then copy the image to the destination region. For more information, see Copy a custom image. Check for issues with your network service provider.
The migration speed depends on the outbound bandwidth of the source server and the inbound bandwidth of the intermediate instance. The bandwidth of the intermediate instance is limited. By default, the maximum inbound public bandwidth is 100 Mbps. Therefore, the maximum transfer speed over the public internet is 100 Mbps by default. Important When you migrate to a destination instance that uses an elastic IP address (EIP), its public inbound bandwidth is limited by the EIP's public outbound bandwidth, as follows: If the outbound bandwidth of the EIP is 10 Mbps or less, the inbound bandwidth of the instance is 10 Mbps. If the outbound bandwidth of the EIP is greater than 10 Mbps, the inbound bandwidth of the instance is, by default, equal to the outbound bandwidth of the EIP.	To resolve this, use one of the following methods: Method 1: In the ECS console, find the intermediate instance or destination instance. Convert the instance's static public IP to a pay-as-you-go elastic IP address (EIP) and increase its peak bandwidth to 200 Mbps. This action incurs charges. For more information, see Convert a static public IP address to an EIP and Modify the bandwidth of an EIP. Important You cannot convert an EIP back to a static public IP. Because EIPs are not released when the associated instance is released, you must manually release the pay-as-you-go EIP after the migration to prevent unexpected charges. For more information, see Release a pay-as-you-go EIP. Method 2: If your source server can access a Virtual Private Cloud (VPC) in an Alibaba Cloud region, select migration over a VPC for your migration job. This method is faster and more stable than migrating over the public internet. You can connect your source server to the VPC by using VPN Gateway, Express Connect, or Smart Access Gateway. For more information, see Connect a VPC to an on-premises data center or another cloud.
Performance bottlenecks on the source server, such as CPU, memory, or disk limitations, can reduce SMC transfer speed.	Address the performance bottlenecks on the source server, for example, by upgrading CPU, memory, or disk performance. Improve transfer efficiency by disabling data compression. In the Create Migration Job > Advanced Settings > Compression Ratio or Edit Job > Compression Ratio section, set the Compression Ratio to 0. Data compression uses CPU resources and is recommended for low-bandwidth scenarios. In high-bandwidth scenarios, it can reduce transfer speeds.
The source server contains large, dynamically changing files, such as database or log files.	Pause related data services during the migration to prevent large files from changing, then retry the migration. Use filter configurations to exclude the relevant directories and then try again. For more information, see How do I filter or exclude files or directories that I do not want to migrate? You can migrate the excluded files or directories later by using other methods.
The source server has a large number of small files, such as data from a distributed file system or a file server.	Use filter configurations to exclude the relevant directories and then try again. For more information, see How do I filter or exclude files or directories that I do not want to migrate? You can migrate the excluded files or directories later by using other methods.
The source server has low outbound bandwidth.	Temporarily increase the outbound bandwidth of the source server. In the Create Migration Job > Advanced Settings > Compression Ratio or Edit Job > Compression Ratio section, enable data compression by setting the Compression Ratio (a value of 7 is recommended). Data compression improves transfer efficiency but consumes some CPU resources. It is not recommended for high-bandwidth scenarios.

Note

The examples in this topic are for reference only.

Estimate migration time

Before you start a server migration or an agentless migration for VMware, you can estimate the time required for a full migration or an incremental migration.

Estimate full migration time

The total time for a full migration is the sum of the data transfer time and the image creation time. The following diagram shows how to estimate the time required. Note the following:

The snapshot creation speed is approximately 100 MB/s.
Automatic snapshot creation typically peaks around midnight. This peak activity can reduce the bandwidth for each snapshot, resulting in longer creation times.

For information about how to test the actual network speed, see Test the data transfer speed.

For example, assume a source server has 10 GB of data in use and an outbound bandwidth of 10 Mbps. The estimated time for a full migration is calculated as follows:

Convert units.
- Actual data volume: 10 GB = 10 × 1,024 = 10,240 MB
- Actual network speed: 10 Mbps = 10/8 = 1.25 MB/s
Calculate the data transfer time.
Data transfer time: 10,240 MB / 1.25 MB/s = 8,192 seconds ≈ 2.27 hours
Calculate the image creation time.
Image creation time: 10,240 MB / 100 MB/s = 102.4 seconds ≈ 0.03 hours
Calculate the total migration time.
Total migration time: 2.27 hours + 0.03 hours = 2.3 hours

Estimate incremental migration time

The time required for an incremental migration consists of the incremental comparison time, data transfer time, and image creation time. The following diagram shows how to estimate the time required. Keep in mind the following:

The snapshot creation speed is approximately 100 MB/s.
Automatic snapshot creation typically peaks around midnight. This peak activity can reduce the bandwidth for each snapshot, resulting in longer creation times.

The incremental comparison speed is approximately 100 MB/s.
This estimate assumes block-level replication is used, where the comparison speed is related to disk I/O performance. If block-level replication is not enabled, the comparison speed depends on the number and size of changed files.
For information about how to test the actual network speed, see Test the data transfer speed.
For agentless migration for VMware, you can enable Changed Block Tracking (CBT), which reduces the incremental comparison time to zero.
For more information, see Incremental agentless migration for VMware.

For example, assume a source server has 40 GB of data in use, 2 GB of incremental data, and an outbound bandwidth of 10 Mbps. The estimated time for an incremental migration is calculated as follows:

Convert units.
- Disk data volume: 40 GB = 40 × 1,024 = 40,960 MB
- Incremental data volume: 2 GB = 2 × 1,024 = 2,048 MB
- Actual network speed: 10 Mbps = 10/8 = 1.25 MB/s
Calculate the incremental comparison time.
Incremental comparison time: 40,960 MB / 100 MB/s = 409.6 seconds ≈ 0.11 hours
Calculate the data transfer time.
Data transfer time: 2,048 MB / 1.25 MB/s = 1,638.4 seconds ≈ 0.46 hours
Calculate the image creation time.
Image creation time: 2,048 MB / 100 MB/s = 20.48 seconds ≈ 0.01 hours
Calculate the total incremental migration time.
Total migration time: 0.11 + 0.46 + 0.01 = 0.58 hours

Test the data transfer speed

The transfer speed from the source server to the intermediate instance depends on both the source server's outbound bandwidth and the intermediate instance's inbound bandwidth.

For example:

If the outbound bandwidth of the source server is 50 Mbps and the inbound bandwidth of the intermediate instance is 100 Mbps, the theoretical maximum transfer speed does not exceed 50 Mbps.
If the outbound bandwidth of the source server is 150 Mbps and the inbound bandwidth of the intermediate instance is 100 Mbps, the theoretical maximum transfer speed does not exceed 100 Mbps.

Note

The 1 Mbps bandwidth displayed for the instance in the ECS console is the outbound bandwidth of the intermediate instance. This does not affect the migration speed, which depends on the inbound bandwidth of the intermediate instance.

Follow these steps to test the transfer speed with the iPerf tool:

Create a pay-as-you-go ECS instance in the target Alibaba Cloud region.
On the instance, do the following:
1. Install the iPerf tool.
2. Start iPerf in server mode.
3. Add a security group rule to open the required iPerf port.
On the source server, do the following:
1. Install the iPerf tool.
2. Start iPerf in client mode. Set the destination server IP address to the public IP address of the pay-as-you-go instance that you created in Step 1.

Linux

These steps use CentOS 7 as an example. The commands may vary depending on your OS version.

Create a pay-as-you-go CentOS 7 instance in the target Alibaba Cloud region.
Add an inbound rule to the instance's security group to allow traffic on the required iPerf port.
In this example, the default iPerf port TCP 5001 is used.
Connect to the CentOS 7 instance.
On the CentOS 7 instance, do the following:
1. Run the following command to install the iPerf tool.
```
yum -y install iperf3
```
2. Run the following command to start iPerf in server mode.
```
iperf3 -s
```
On the source server, do the following:
1. Download and install the iPerf tool.
2. Run the following command to start iPerf in client mode.
  Replace <instance_ip> with the public IP address of the instance that you created.
```
iperf3 -c <instance_ip> -i 1 -d
```
Wait for the test to complete and record the results.

Windows

These steps use Windows Server 2008 as an example. The steps may vary depending on your OS version.

Create a pay-as-you-go Windows Server 2008 instance in the target Alibaba Cloud region.
Add an inbound rule to the instance's security group to allow traffic on the required iPerf port.
In this example, the default iPerf port TCP 5001 is used.
Connect to the instance.
On the Windows Server 2008 instance, do the following:
1. Download and install the iPerf tool.
2. Open Command Prompt.
3. Run the cd <iperf_directory> command to navigate to the tool's directory.
4. Run iperf3.exe -s to start iPerf in server mode.
On the source server, do the following:
1. Download and install the iPerf tool.
2. Run the following command to start iPerf in client mode.
  Replace <instance_ip> with the public IP address of the instance that you created.
```
iperf3.exe -c <instance_ip> -i 1 -d
```
Wait for the test to complete and record the results.