For customers that have enabled IBM’s call home feature, data is reported back to IBM that is used not only for service requests but also helps IBM track usage and trends of various features. Live Partition Mobility (LPM) provides quite a bit of usage information that is reviewed by IBM. This blog will give some insights into the trends that IBM is seeing with respect to LPM usage.
LPM Success Rate
The key metric that IBM reviews is the success rate of LPM operations in the field. Since LPM was introduced in POWER6, the success rate has continued to improve on a yearly basis.
Year
|
LPM Success Rate
|
2020
|
96.41%
|
2021
|
97.21%
|
2022
|
97.60%
|
Year To Date 2023
|
97.72%
|
In addition to the success rate, the LPM Call Home data includes reason codes that help to identify the reason for the LPM failure. IBM reviews these reason codes such that changes can be made to LPM validation or updates to error messages to reduce the time required for customers to react to these failures. The most common reasons for LPM failures are related to zoning and masking issues for the SAN storage. One way to ensure that both WWPNs are setup correctly is to leverage the –-npivval parameter on the migrlpar HMC command. This can be configured to not only verify the port is accessible from the secondary WWPN but also the individual LUNs. It’s a good business practice to perform an LPM validation operation periodically and after making configuration changes to the SAN just to make sure everything is still configured correctly to reduce downtime in situations where you need to perform LPM.
OS Migration Rates
Originally when LPM was developed, it was only supported for AIX. After the initial introduction, support was added for IBM i and later Linux. Since 2020 to now, the percentage of IBM i and Linux requests has doubled. Due to the applications that are supported on Power Linux, the call home data shows there are numerous LPM operations that have been reported this year that exceed 10 TB of memory.
Effective Network Speeds
The LPM Call home data reports a measurement of effective network speed of the LPM operation. This is a combined measure of the speed of hardware in the network coupled with compression of the data by the PowerVM Hypervisor which leverages the built-in hardware compression engines. Compression and Encryption of the data was added starting with FW920 on POWER9. The following table shows a breakdown of the effective network speeds over time.
Year
|
<=1Gb
|
>1Gb and <=10Gb
|
>10Gb and <=40Gb
|
>40Gb
|
2020
|
16.71%
|
64.37%
|
18.83%
|
0.09%
|
2021
|
13.96%
|
68.34%
|
17.62%
|
0.08%
|
2022
|
10.33%
|
66.11%
|
23.47%
|
0.09%
|
Year To Date 2023
|
10.51%
|
69.48%
|
19.97%
|
0.04%
|
Based on the data, customers are migrating away from 1Gb network connections and utilizing higher speed networks for LPM operations. The general LPM guidelines is to utilize 10Gb or better dedicated connections but for large memory configurations or to achieve better LPM performance, customers should be planning for improvements in hardware network speeds.
Time Required to Perform LPM Operations
The following table is a summary of the LPM call home data summarizing the time it is taking to migrate the partition data from the source to the target server:
Year
|
<=5 minutes
|
>5 minutes and <=10 minutes
|
>10 minutes and <= 15 minutes
|
> 15 minutes
|
2020
|
87.90%
|
7.34%
|
2.16%
|
2.6%
|
2021
|
91.93%
|
5.10%
|
1.47%
|
1.51%
|
2022
|
92.53%
|
4.57%
|
1.44%
|
1.46%
|
Year To Date 2023
|
93.36%
|
4.21%
|
1.19%
|
1.25%
|
One item to note is that 97% of the partition migrations are completing in 10 minutes or less.
LPM times are a combination of network speed and partition memory sizes. The following table provides insights into the amount of data that needs to be transmitted for LPM requests:
Year
|
<=5 GB
|
>5GB and <=20 GB
|
>20GB and <=50GB
|
>50GB and <=100GB
|
>100GB
|
2020
|
20.95%
|
42.74%
|
23.16%
|
8.08%
|
5.07%
|
2021
|
33.82%
|
35.92%
|
18.58%
|
7.16%
|
4.53%
|
2022
|
29.02%
|
38.47%
|
19.97%
|
7.28%
|
5.26%
|
Year To Date 2023
|
32.64%
|
18.29%
|
18.29%
|
6.41%
|
4.94%
|
The reduction in the amount of memory that needs to be transmitted is directly proportional to the automatic compression of the data along with higher network speeds (faster networks reduce the amount of changed data that needs to be transmitted).
Summary
The IBM Call Home support is a feature that is leveraged quite a bit by the hardware teams and the LPM teams. The coverage provided by call home is not limited to just LPM, there are many features that are tracked which helps IBM improve and develop future products and services. You can find more information about the IBM Call Home data at How does PowerVM Development use your Call Home Data .
Contacting the PowerVM Team
Have questions for the PowerVM team or want to learn more? Follow our discussion group on LinkedIn IBM PowerVM or IBM Community Discussions